The Journal of Nutrition

Nutrient Requirements and Optimal Nutrition

Nondigestible Carbohydrates Affect Metabolic

Health and Gut Microbiota in Overweight
Adults after Weight Loss
Alexandra M Johnstone,1 Jennifer Kelly,2 Sheila Ryan,2 Reyna Romero-Gonzalez,1 Hannah McKinnon,1
Claire Fyfe,1 Erik Naslund,3 Ruben Lopez-Nicolas,4 Douwina Bosscher,5 Angela Bonnema,6
Carmen Frontela-Saseta,4 Gaspar Ros-Berruezo,4 Graham Horgan,7 Xiaolei Ze,1 Jo Harrold,8
Jason Halford,8 Silvia W Gratz,1 Sylvia H Duncan,1 Soraya Shirazi-Beechey,2 and Harry J Flint1

1
The Rowett Institute, University of Aberdeen, Foresterhill, United Kingdom; 2 Functional and Comparative Genomics, and Psychological
Sciences, University of Liverpool, Liverpool, United Kingdom; 3 Karolinska Institute, Stockholm, Sweden; 4 Department of Food Science and
Nutrition, Faculty of Veterinary Sciences, Regional Campus of International Excellence “Campus Mare Nostrum”, University of Murcia,
Murcia, Spain; 5 Cargill R&D Centre Europe, Vilvoorde, Belgium; 6 Cargill R&D Centre NA, Minneapolis, MN, USA; 7 Biomathematics
and Statistics Scotland, Aberdeen, United Kingdom; and 8 Appetite and Obesity Research Group, Department of Psychological Sciences,
University of Liverpool, Liverpool, United Kingdom

ABSTRACT
Background: The composition of diets consumed following weight loss (WL) can have a significant impact on satiety
and metabolic health.
Objective: This study was designed to test the effects of including a nondigestible carbohydrate to achieve weight
maintenance (WM) following a period of WL.
Methods: Nineteen volunteers [11 females and 8 males, aged 20–62 y; BMI (kg/m2 ): 27–42] consumed a 3-d
maintenance diet (15%:30%:55%), followed by a 21-d WL diet (WL; 30%:30%:40%), followed by 2 randomized 10-
d WM diets (20%:30%:50% of energy from protein:fat:carbohydrate) containing either resistant starch type 3 (RS-
WM; 22 or 26 g/d for females and males, respectively) or no RS (C-WM) in a within-subject crossover design without
washout periods. The primary outcome, WM after WL, was analyzed by body weight. Secondary outcomes of fecal
microbiota composition and microbial metabolite concentrations and gut hormones were analyzed in fecal samples and
blood plasma, respectively. All outcomes were assessed at the end of each dietary period.
Results: Body weight was similar after the RS-WM and C-WM diets (90.7 and 90.8 kg, respectively), with no difference
in subjectively rated appetite. During the WL diet period plasma ghrelin increased by 36% (P < 0.001), glucose-dependent
insulinotropic polypeptide (GIP) decreased by 33% (P < 0.001), and insulin decreased by 46% (P < 0.001), but no
significant differences were observed during the RS-WM and C-WM diet periods. Fasting blood glucose was lower after
the RS-WM diet (5.59 ± 0.31 mmol/L) than after the C-WM diet [5.75 ± 0.49 mmol/L; P = 0.015; standard error of the
difference between the means (SED): 0.09]. Dietary treatments influenced the fecal microbiota composition (R2 = 0.054,
P = 0.031) but not diversity.
Conclusions: The metabolic benefits, for overweight adults, from WL were maintained through a subsequent WM
diet with higher total carbohydrate intake. Inclusion of resistant starch in the WM diet altered gut microbiota composition
positively and resulted in lower fasting glucose compared with the control, with no apparent change in appetite. This
trial was registered at clinicaltrials.gov as NCT01724411. J Nutr 2020;150:1859–1870.

Keywords: gut microbiota, gut health, resistant starch, weight loss, obesity, appetite, fiber, non-digestible
carbohydrate, human

Introduction carbohydrate have been widely adopted for WL, as they appear
to contribute to satiety (1, 2). However, such diets also lead
Obesity is a major worldwide health problem and is associated
to an altered intestinal metabolic profile, resulting in increased
with increased risk of noncommunicable diseases. Thus, there
concentrations of hazardous nitrogenous metabolites, and if
is a need for dietary interventions to assist with weight loss
consumed over lengthy periods, might compromise gut health
(WL) and facilitate weight maintenance (WM) following a
(3). Thus, it is essential to identify alternative dietary approaches
period of WL. Diets with high protein and decreased total
that assist with maintenance of a healthy weight.

Copyright  C The Author(s) on behalf of the American Society for Nutrition 2020.
Manuscript received August 21, 2019. Initial review completed October 16, 2019. Revision accepted April 9, 2020.
First published online June 8, 2020; doi: https://doi.org/10.1093/jn/nxaa124. 1859
 Digestible carbohydrates are hydrolyzed in the small intes- with obesity and comorbidities (17). It has been shown that
tine by the host intestinal enzymes to monosaccharides such changes in diet alter the composition of the gut microbiota (18,
as glucose. However, nondigestible (ND) carbohydrates present 19). Dietary approaches, designed to enhance the abundance
in most diets in the form of plant fiber and resistant starch and activity of beneficial gut bacteria in order to improve and
cannot be broken down in the small intestine and reach the large maintain healthy gut function, are of growing importance for
intestine (colon) where they are fermented by the gut microbiota human health (20).
to SCFAs, which are absorbed by specific transporters across This protocol, registered at clinicaltrials.gov (ID:
the colonic epithelial cells (4, 5). Since starch fermentation to NCT01724411), was conducted to investigate the effect
SCFAs provides considerably less energy per gram compared of ND fibers on the primary outcome: WM after WL. This
with intake of the same amount of digestible carbohydrate (6), was linked to the secondary outcomes: to assess the impact
it results in a lower glycemic index. There is also evidence of the fibers on potential mechanisms via the gut microbiota
that ND carbohydrates can improve satiety via effects on gut (as measured by SCFAs and qPCR) and plasma gut hormone
function through microbially produced propionate and butyrate profiles (fasted and postprandial glucose, insulin, lipids, and
(7). These SCFAs are sensed by specific G protein–coupled gut peptides) in healthy but obese and overweight volunteers in
receptors expressed in intestinal endocrine cells eliciting the free-living conditions.
release of satiety hormones, glucagon-like peptide 1 (GLP- Exploratory outcomes were the anthropometric variables
1) and peptide YY (PYY). These gut hormones influence BMI, resting metabolic rate (RMR), % fat mass, % fat-free
appetite by signaling to the hypothalamus through the afferent mass, fat mass (kilograms), fat-free mass (kilograms), waist
nerves (8). circumference, hip circumference, waist-to-hip ratio (WHR),
Resistant starch (RS) provides a potentially convenient systolic (SBP) and diastolic (DBP) blood pressure, and pulse;
source of ND carbohydrate that has also been reported to daily food intake during each dietary period; and subjective
improve insulin sensitivity (9, 10). The metabolic link between appetite ratings (hunger, fullness, desire to eat, amount intended
RS ingestion and insulin sensitivity is unclear, but may be to eat) assessed with visual analog scales (VASs).
through alterations in SCFA production (11). Being a source of
dietary fiber, RS offers potential health benefits on the digestive
tract and capacity to alter the gut microbiota profile, but it
has also been linked to improvements in blood cholesterol Methods
profile and lower glycemic index [see Lockyer and Nugent Subject recruitment
(12) for a review]. Commercial preparations can be added to Recruitment of volunteers [BMI (kg/m2 ): 27–42] was by public
foods as a functional ingredient and offer a way to increase advertisement of a diet trial. Inclusion criteria specified that volunteers
the fiber content of food, without decreasing the textural and should not have existing medical conditions or medications that could
organoleptic characteristics (13). RS type 3 (RS3), in particular, influence their appetite or mood, and this was confirmed by a medical
offers advantages as a heat-stable prebiotic ingredient in a examination. We screened 40 subjects, and recruited 24 to enter the
variety of conventional foods (14). study to be randomly assigned to a diet, with 19 subjects completing
the study (11 females and 8 males). [See the CONSORT (Consolidated
The connection between changes in gut microbiota, inflam-
Standards of Reporting Trials) flow diagram (Figure 1) summarizing the
mation, and the pathogenesis of obesity-related disorders has
participant flow.]
been increasingly recognized (15). Gut bacteria can initiate the We were unable to collect blood samples from 1 subject, so all
inflammatory state of obesity and insulin resistance through the blood parameter analysis is presented for n = 18. All subjects gave
activity of LPS and their stimulating role on macrophages and written, informed consent. The study was approved by the North of
lymphocytes to secrete inflammatory cytokines including TNF- Scotland Research Ethics Service. All the participants visited the Human
α and IL-6 (16). Also, SCFAs may exert modulatory effects on Nutrition Unit at the Rowett Institute (Aberdeen, UK). Dietetic staff
inflammatory markers, thus contributing to the link between supplied all food and drink consumed during all dietary periods, and
gut microbiota and low-grade inflammatory state associated participants attended 3 times/wk for body-weight assessment and food
provision. The protocol was a within-subject crossover design and
lasted 49 d, as described in the protocol diagram shown in Figure 2,
The research leading to these results has received funding from the as follows: days 1–4, habitual diet consumed ad libitum (4 d); days 5–7,
European Community’s Seventh Framework Program FP7/2007–2013 under
maintenance diet (M; 3 d); days 8–28, high-protein WL diet (WL; 21 d);
grant agreement no. KBBE- 2011–5-289800.
Author disclosures: AB and DB are employed by Cargill, Inc., which is the
days 29–38 and 39–48, randomly assigned WM diets (10 d), RS (RS-
producer of Actistar resistant starch. The other authors report no conflicts WM) or control (C-WM). Subjects were provided with a breakfast test
of interest. Authors from the University of Aberdeen, The Rowett Institute, meal on 4 occasions, at the end of each dietary phase, corresponding to
gratefully acknowledge financial support from the Scottish Government as part the morning of day 8, 29, 39, and 49.
of the Strategic Research Programme at the Rowett Institute (start date 1 April,
2016 – 31 March, 2022) for technical staff input.
Supplemental Tables 1–7, Supplemental Methods, and Supplemental Figure 1 Formulation and preparation of the diets
are available from the “Supplementary data” link in the online posting of the The study diets were designed with protein, fat, and carbohydrate
article and from the same link in the online table of contents at https://academ content expressed as a percentage of total energy. The M diet (days 5–
ic.oup.com/jn/. 7) consisted of 15% protein, 30% fat, and 55% carbohydrate fed to
Address correspondence to AMJ (e-mail: alex.johnstone@abdn.ac.uk). 1.5 × measured RMR, applied to reflect UK average nutrient intake
Abbreviations used: BCFA, branched-chain fatty acid; C-WM, control weight- (21). The WL diet was fed as 100% RMR as a 7-d rotation menu,
maintenance; DBP, diastolic blood pressure; EGP, endogenous glucose produc-
as 3 meals/d, each 30% protein, 30% fat, and 40% carbohydrate.
tion; GIP, glucose-dependent insulinotropic polypeptide; GLP-1, glucagon-like
peptide 1; iAUC, incremental AUC; M, maintenance; ND, nondigestible; OTU,
During the subsequent periods either control (C-WM) or RS (RS-
operational taxonomic unit; PYY, peptide YY; RMR, resting metabolic rate; RS, WM) were provided as 1.2 × RMR with 20% protein, 30% fat,
resistant starch; RS3, resistant starch type 3; RS-WM, resistant-starch weight- and 50% carbohydrate, designed with slightly higher protein intake
maintenance diet; SBP, systolic blood pressure; SED, standard error of the to maintain compliance to the diet and fed with slightly less energy
difference between the means; tAUC, total AUC; VAS, visual analog scale; WHR, than baseline to reflect their lower requirements due to WL. In the RS-
waist-to-hip ratio; WL, weight loss; WM, weight maintenance. WM diet, a daily amount of 22 g for females and 26 g for males of

1860 Johnstone et al.

FIGURE 1 CONSORT diagram summarizing participant flow with the sizes (n) of initial (recruited, enrolled) and final groups. CONSORT,
Consolidated Standards of Reporting Trials.

RS3 (C∗ActiStar 11700; Cargill) was provided, and a digestible starch kit; Megazyme International). A sample menu for each diet phase is
was incorporated as control in the C-WM diet. A 5-d rotational menu provided in Supplemental Tables 1–5. In general, there are 4 different
was created for these dietary periods, and all meals were the same types of RS and this RS3 was chosen for practical reasons due to
weight and isocaloric. The composition of each meal, in terms of energy, its ability to be incorporated into a wide variety of food matrices, in
fat, carbohydrate, and protein, was calculated by using McCance and particular being baked into bread while still maintaining a desirable
Widdowson’s The Composition of Foods (22). Energy and nutrient texture; it is a retrograded starch that is created when starch is cooked
intakes were calculated by using NETWISP™ software (version 3.0; and cooled, and the glucose chains are released upon cooking but some
Tinuviel). RS3 was incorporated into a wide range of matrices (i.e., of them crystallize and form bonds that slow down digestion. Other
bread, biscuits), with a percentage of RS incorporation from 10% to food forms include cooked and cooled potatoes, rice, and pasta, which
29% RS3, determined using the AOAC method 2002.02 (K-RSTAR were also incorporated into the diet to maintain palatability.

FIGURE 2 Experimental design for overweight adults in a 49-d dietary protocol. n = 19; 11 males and 8 females; aged 20–62 y; BMI ( kg/m2 ):
27–40. C-WM, control weight-maintenance diet; M, maintenance diet; RS-WM, resistant-starch weight-maintenance diet; WL, weight-loss diet.

Impact of complex carbohydrates on gut microbiota 1861

Measurement of anthropometric variables, RMR, and numbers were used to assign the subjects. Data on energy intake,
blood pressure body weight, body composition, and blood metabolites, presented
Measurements of body composition and metabolic rate were conducted as end-of-treatment values, were analyzed by ANOVA, with subject
at the end of each diet period under standardized conditions described as blocking factors (random effect) and diet (M, WL, C-WM, RS-
previously (23), including height and weight, 2-compartment model of WM) as treatment term (fixed effect). When the effect of diet was
body composition by air-displacement whole-body plethysmography, significant (P < 0.05), means were compared with Fisher’s least-
abdominal and gluteal (hip) circumference, and blood pressure. RMR significant-difference post hoc t test. Data on appetite ratings were
rate was measured using indirect calorimetry (Quark RMR; COSMED), analyzed by mixed models, with random-effect terms for subject, period
taken before and after each dietary period to assess changes in energy within subject, day within period, and hour within day and fixed-effect
requirements. terms for diet, day of intervention, time of day (by clock time), and
their interactions. Individual VAS ratings were a dependent variable.
Assessment of appetite and pleasantness of meals Normality was assessed by inspection of histograms of variables and
Subjective appetite was assessed hourly (by clock time) during the last 3 residuals. If variables showed some slight indications of skewness,
d of each dietary period and every 30 min (from time zero) during test analysis was repeated on a log scale to confirm that conclusions
days for 180 min with 100-mm VASs, as detailed previously (24). remained unchanged. All analyses were carried out using Genstat
version 13 (Lawes Agricultural Trust, VSN International Ltd.). To
investigate associations between genus-level microbiota abundance
Compliance, metabolic profile, inflammatory markers,
and bodily parameters, Spearman rank correlation analysis was
and gut hormone peptide measurement
undertaken.
Subjects consumed a test meal on days 8, 29, 39, and 49. On average,
meals provided one-third of daily energy requirements, and each meal
was identical in composition within each diet phase. Before and after
the meal, blood samples were collected every 30 min for 3 h from a
cannula in the arm. The samples were centrifuged at 4◦ C for 10 min, Results
at 1000 × g, with the plasma placed in aliquots, and stored at −70◦ C
until analysis in 1 single run. Subject characteristics
All blood samples were collected in S-Monovette® (Sarstedt) tubes. Nineteen subjects completed the study and their baseline
Plasma glucose and lipid profiles were analyzed using KONE (Espoo), anthropometric characteristics are presented in Table 1.
as described previously (1). For gut peptide sampling (PYY, ghrelin, GIP,
and GLP-1), equal parts of DiPeptidyl Peptidase IV soluble inhibitor
(Merck Millipore), protease inhibitor cocktail (Sigma Aldrich Co. LLC), Impact of dietary intervention on food intake, body
and Pefabloc SC® —ABSF (Roche) were added, and the plasma was weight, and body composition
analyzed using Milliplex Human metabolic hormone panel bead-based
Participant mean habitual energy intake was 2110 ± 641 kcal/d
immunoassay (Millipore Corp.).
before the study, as recorded by free-living food diaries. There
Inflammatory biomarkers (IL-8, IL-1β, IL-6, IL-10, TNF-α, and IL-
12p70) were analyzed in plasma samples within 3 mo of collection was no significant difference in energy intake between the C-
with the CBA Human Inflammatory Cytokine Kit (Becton Dickinson WM and RS-WM diet periods (Table 2). Mean WL during the
Biosciences) (25). Data acquisition was performed using FACS Calibur® study was 3.64 kg ± 1.93 (range: 0.65–7.15 kg). The significant
flow cytometer (Becton Dickinson Biosciences) and BD Bioscience CBA reduction in body weight after the WL period, relative to
software for data analysis. baseline, was maintained after the C-WM and RS-WM diet
C-reactive protein (CRP) was measured in plasma using a high- periods (P < 0.001; standard error of the difference between
sensitive immunoturbidimetric assay (1–160 mg/L; Kamiya Biomedical the means, SED: 0.32) (Table 3). No difference was observed
Company). after the C-WM and RS-WM diet periods, and subjects were
weight stable with the reduced maintenance diet provision.
Fecal sample DNA extraction and qPCR and 16S ANOVA of body composition indicated no statistical
ribosomal RNA amplicon sequencing difference for loss of lean mass and fat mass between diets
Fecal samples obtained from all subjects on study days 5, 8, 26, 29, 37,
(P = 0.30; SED: 0.81). SBP was reduced significantly by 6%
39, 47, and 49 were stored at 4◦ C and processed within 5 h. Full details,
statistical methods, and references for the fecal sample DNA extraction after the WL diet and was maintained after both WM diets
and qPCR and subsequent 16S ribosomal RNA amplicon sequencing (P < 0.001; SED: 2.07). A similar effect was observed for
are provided as Supplemental Methods. Illumina sequence data has been DBP, where a 5% reduction occurred after the WL diet and
deposited in the European Nucleotide Archive under project accession was maintained after the C-WM and RS-WM diet periods
number PRJEB18768. (P < 0.001; SED: 1.75). Reductions of 2.57 cm in waist
circumference and 2.35 cm in hip circumference occurred after
Bacterial metabolite analysis the WL diet period and decreased by another 1 cm during the C-
The SCFA content of fecal samples was determined by capillary GC WM and RS-WM diet periods (P = 0.006; SED: 1.08 for waist
following conversion to t-butyldimethylsilyl derivatives (26). The lower circumference; P = 0.014; SED: 0.87 for hip circumference).
limit for reliable detection of each product was taken as 0.2 mM. N- Despite these findings, no change was observed in WHR
nitroso compounds were measured in fecal water (from aliquots stored (P = 0.17; SED: 0.01).
at −70◦ C), as described previously (3).

Tolerability and safety of the diets Motivation to eat and pleasantness of the diets
Subjective gastrointestinal responses were monitored during the last
ANOVA performed using 3 d VAS data confirmed that, across
3 d of each dietary intervention period using a previously validated
questionnaire (27).
the study, subjectively rated hunger (P = 0.50; SED: 10.9),
fullness (P = 0.50; SED: 9.06), desire to eat (P = 0.29; SED:
9.89), and amount intended to eat (P = 0.56; SED: 10.6) were
Statistical analysis of bodily parameters
A sample size of 20 subjects has ∼90% power in a 2-sided test, at α not different between dietary interventions (data not shown).
= 0.05 in unadjusted tests to detect differences comparable to within- Palatability was highly rated for all meals (mean range: 78–89
subject variability (Cohen’s d = 1.0). Computer-generated random out of 100 mm on the VAS).

1862 Johnstone et al.

TABLE 1 Baseline anthropometric characteristics of overweight adults who completed a 49-d dietary protocol1

Gender
Variable Males (n = 11) Females (n = 8) SED All (n = 19)
Weight, kg 110 ± 14.1 83.1 ± 12.3 0.47 94.3 ± 18.5
BMI, kg/m2 34.4 ± 4.65 31.7 ± 3.47 0.17 32.8 ± 4.07
Fat mass, % 33.2 ± 5.65 42.5 ± 5.90 1.19 38.6 ± 7.35
Fat-free mass, % 66.8 ± 5.65 57.5 ± 5.90 1.19 61.4 ± 7.35
Fat mass, kg 36.3 ± 7.93 35.7 ± 8.80 1.28 36.0 ± 8.22
Fat-free mass, kg 72.6 ± 8.26 47.4 ± 5.78 1.14 58.0 ± 14.4
Waist circumference, cm 113 ± 12.8 95.7 ± 12.3 1.57 103 ± 14.9
Hip circumference, cm 114 ± 8.44 109 ± 7.66 1.27 112 ± 8.08
WHR 1.00 ± 0.06 0.86 ± 0.08 0.01 0.92 ± 0.10
SBP, mm Hg 141 ± 13.5 128 ± 12.7 3.02 133 ± 14.3
DBP, mm Hg 80.8 ± 7.19 79.7 ± 8.50 2.29 80.2 ± 7.78
Pulse, bpm 70.2 ± 7.92 71.9 ± 10.67 2.37 71.2 ± 9.40
1
Values are means ± SDs. bpm, beats per minute; DBP, diastolic blood pressure; SBP, systolic blood pressure; SED, standard error of the difference between the means; WHR,
waist-to-hip ratio.

Influence of diet on plasma lipid profiles and glucose Triglycerides were also significantly lower after the C-WM and
tolerance RS-WM diet periods compared with after consumption of the M
As anticipated, mean fasting glucose significantly decreased diet (Table 4). Nonesterified fatty acids were significantly lower
by 3.8% after the WL diet period and further decreased by after the C-WM and RS-WM diet periods compared with after
0.5% after RS-WM (P = 0.015; SED: 0.09) relative to after the WL diet period (P = 0.015; SED: 0.04).
the C-WM diet period (Table 4). Similarly, total cholesterol Table 5 outlines the data related to the postprandial response
was significantly reduced by 9.8% after the WL diet period to the test meal for glucose and insulin. There was no effect
(P < 0.001; SED: 0.109), and this reduction was maintained on the 180-min plamsa glucose profile due to diet, expressed
after the C-WM and RS-WM diet periods, along with a as total AUC (tAUC) or incremental AUC (iAUC). There was,
significant reduction in LDL cholesterol that was maintained however, a large and significant effect on plasma insulin due
during both WM dietary periods (P = 0.003; SED: 0.082). We to diet, with an expected decrease after the WL diet period
also observed a 33.3% reduction in mean fasting triglycerides relative to after the M diet period expressed as both tAUC
associated with WL (P < 0.001; SED: 0.007) (Table 4). (−46%; P < 0.001; SED: 76.6) and iAUC (−49%; P < 0.001;

TABLE 2 Macronutrient composition of the daily food consumed by overweight adults during the study M, WL, C-WM, and RS-WM
dietary periods1

Variable M WL C-WM RS-WM SED P2

Energy, MJ 11.4c ± 2.38 8.05a ± 1.87 9.22b ± 2.60 9.36b ± 2.73 255 <0.001
Energy, kcal 2730c ± 569 1930a ± 446 2200b ± 623 2240b ± 653 60 <0.001
Energy density, kJ/100 g 367c ± 97 267a ± 95 272b ± 96 274b ± 99 13 <0.001
Fat, % 29.8a ± 1.24 29.8a ± 1.58 30.0a ± 1.94 30.3a ± 2.27 0.50 <0.001
Protein, % 15.1a ± 1.07 29.4c ± 1.98 20.1b ± 1.63 20.0b ± 1.52 0.30 <0.001
Carbohydrate, % 54.6c ± 2.84 40.0a ± 2.28 49.5b ± 2.66 49.3b ± 2.69 40.00 <0.001
Saturated fat, % 10.8a ± 2.20 10.3a ± 1.62 11.4b ± 1.96 12.2c ± 2.06 0.30 <0.001
Salt, g 9.29d ± 2.13 8.68c ± 2.20 8.32a,b ± 2.42 8.20a ± 2.52 0.30 <0.001
Alcohol, kJ 39a ± 289 77a ± 249 90a ± 389 68a ± 352 51 <0.001
Sugar, g 179c ± 44.0 74.8a ± 23.2 111b ± 44.9 105b ± 42.2 5.10 <0.001
Starch, g 208d ± 43.9 117a ± 30.3 167c ± 46.8 159b ± 45.2 5.00 <0.001
Sugar:starch ratio 0.87b ± 0.14 0.66a ± 0.16 0.67a ± 0.25 0.67a ± 0.26 0.10 <0.001
NSP, g 27.2c ± 7.07 17.5a ± 4.90 17.8a ± 6.94 19.8b ± 6.47 0.90 <0.001
NMES, g 99.6c ± 34.4 44.5a ± 14.8 63.2b ± 28.3 60.9b ± 28.8 4.00 <0.001
Dietary fiber, g 39.9c ± 9.61 24.8b ± 6.85 24.0a ± 9.44 41.3c ± 12.1 1.20 <0.001
Insoluble fiber, g 21.6c ± 5.79 14.9b ± 4.04 10.0a ± 4.17 10.0a ± 4.32 0.50 <0.001
Soluble fiber, g 7.70c ± 2.12 5.17b ± 1.43 3.64a ± 1.43 3.67a ± 1.42 0.20 <0.001
Insoluble:soluble fiber ratio 2.82 ± 0.48 2.87 ± 0.32 2.79 ± 0.49 2.80 ± 0.58 0.10 0.126
C∗Actistar 11700 powder (Cargill), g — — — 32.5 ± 8.80 0.60 <0.001
Resistant starch, g 12.7c ± 3.18 7.25b ± 2.14 6.20a ± 3.10 21.5d ± 6.19 0.40 <0.001
1
Values are means ± SDs, n = 19. Means in the same row not sharing a superscript letter differ (P < 0.05). C-WM, control weight-maintenance diet; M, maintenance diet;
NMES, nonmilk extrinsic sugars; NSP, nonstarch polysaccharides; RS-WM, resistant-starch weight-maintenance diet; SED, standard error of the difference between the means;
WL, weight-loss diet; —, no values. Data were normally distributed.
2
Analyzed by ANOVA.

Impact of complex carbohydrates on gut microbiota 1863

TABLE 3 Biometric data of overweight adults after consumption of M, WL, C-WM, and RS-WM diets (study days 8, 29, 40, and 50)1

Variable M WL C-WM RS-WM SED P2

Weight, kg 94.3c ± 18.5 91.5b ± 17.8 90.7a ± 18.0 90.8a ± 17.8 0.32 <0.001
BMI, kg/m2 32.8c ± 4.07 31.8b ± 3.83 31.6a ± 3.83 31.6a ± 3.74 0.11 <0.001
Fat mass, % 38.6 ± 7.35 37.9 ± 7.36 37.1 ± 7.76 37.5 ± 7.36 0.81 0.303
Fat-free mass, % 61.4 ± 7.35 62.2 ± 7.36 62.9 ± 7.76 62.5 ± 7.36 0.81 0.303
Fat mass, kg 36.0b ± 8.22 34.7ab ± 9.38 33.4a ± 8.40 33.9a ± 8.08 0.87 0.028
Fat-free mass, kg 58.0 ± 14.4 57.2 ± 13.8 57.7 ± 15.3 57.1 ± 14.5 0.78 0.646
Waist circumference, cm 103b ± 14.9 100a ± 15.2 99.5a ± 14.8 99.0a ± 14.9 1.08 0.006
Hip circumference, cm 112b ± 8.08 109a ± 8.01 109a ± 8.37 109a ± 7.82 0.87 0.014
WHR 0.92 ± 0.10 0.91 ± 0.11 0.91 ± 0.11 0.90 ± 0.11 0.01 0.170
SBP, mm Hg 133b ± 14.3 126a ± 14.0 123a ± 12.1 126a ± 15.3 2.07 <0.001
DBP, mm Hg 80.2b ± 7.78 75.3a ± 7.30 77.3ab ± 9.54 75.8a ± 7.28 1.57 0.004
Pulse, bpm 71.2b ± 9.40 69.1ab ± 6.92 67.6a ± 8.53 69.7 ab ± 8.25 1.62 0.002
1
Values are means ± SDs, n = 19. Means in the same row not sharing a superscript letter differ (P < 0.05). bpm, beats per minute; C-WM, control weight-maintenance diet;
DBP, diastolic blood pressure; M, maintenance diet; RS-WM, resistant-starch weight-maintenance diet; SBP, systolic blood pressure; SED, standard error of the difference
between the means; WHR, waist-to-hip ratio; WL, weight-loss diet. Data were normally distributed.
2
Analyzed by ANOVA.

SED: 68.6). This effect was maintained after the RS-WM diet 1811 bacterial operational taxonomic units (OTUs). Phylum-
period but increased by 31% and 36% after the C-WM diet level analysis of microbial composition revealed that Firmicutes
compared with after the WL diet period for tAUC and iAUC, dominated the microbiota, accounting for 72.5% ± 10.7% of
respectively. the total community, followed by Bacteroidetes (17% ± 9.1%)
and Actinobacteria (6.7% ± 8.1%) (Supplemental Figure 1A).
Effect of diet on fecal metabolites The most abundant bacterial genera were Faecalibacterium
There was no significant impact of diet upon acetate, propi- (12.3% ± 6%), Ruminococcus (8.9% ± 6.3%), Bacteroides
onate, or butyrate in fecal samples when expressed as % of (8.4% ± 5.6%), Blautia (7.9% ± 4.4%), Bifidobacterium
total SCFAs (Figure 3A). However, 2 branched chain fatty acids (6.7% ± 8.2%), and Roseburia (5.5 ± 4.3%) (Supplemental
(BCFAs), iso-valerate and iso-butyrate, showed significantly Figure 1B, Figure 4), consistent with previous studies (18, 28,
higher proportions after the WL and C-WM diet periods 29).
compared with after the M and RS-WM diet periods (P < 0.05) The subjects in this study exhibited a high level of
(Figure 3B). Potentially carcinogenic N-nitroso compounds in interpersonal variation, which accounted for 61.3% of the total
fecal water were not significantly different after the 4 diet intersample variation between microbiota profiles (ADONIS
periods (mean ± SD; M: 1040 ± 791 ng/mL; WL: 839 ± 756; software database, BOK group, R2 = 0.61, P < 0.001). As
C-WM: 962 ± 1290; RS-WM: 884 ± 839 ng/mL). There visualized by nonmetric multidimensional scaling analysis, it
were marked differences in total SCFA concentrations across is apparent that samples were grouped mostly by subject
the volunteers. It is recognized that there is considerable rather than diet (Figure 4A). Dietary changes were also found
interindividual variation in gut microbiota composition and to significantly influence composition of the fecal microbiota
gut transit, which impacts on total SCFA concentrations, and (ADONIS software, R2 = 0.054, P = 0.031), particularly “diet
therefore we present data on the main SCFA proportions. responsive” bacterial species, rather than the entire community;
therefore, diet did not impact microbial diversity and/or richness
Dietary effect on fecal microbiota composition as measured by Chao1 and Shannon indices (P > 0.05).
Composition of the fecal microbiota was determined using Each dietary stage had a significant influence upon microbial
high-throughput Illumina sequencing and yielded >47 million groups. The addition of RS caused distinct changes by targeting
high-quality reads, which were subsequently clustered into bacterial groups mainly belonging to the genera Roseburia,

TABLE 4 Fasted plasma glucose and lipid profiles in overweight adults after consumption of M, WL, C-WM, and RS-WM diets
(study days 8, 29, 40, and 50)1

Variable M WL C-WM RS-WM SED P2

Glucose, mmol/L 5.84c ± 0.57 5.62a,b ± 0.58 5.75b,c ± 0.49 5.59a ± 0.31 0.09 0.015
Insulin, pg/mL 747 ± 373 614 ± 228 702 ± 260 752 ± 395 89 0.162
Total cholesterol, mmol/L 4.91b ± 0.83 4.43a ± 0.87 4.63a ± 1.08 4.53a ± 0.99 0.11 <0.001
HDL cholesterol, mmol/L 1.13 ± 0.31 1.07 ± 0.28 1.08 ± 0.30 1.08 ± 0.28 0.32 0.222
LDL cholesterol, mmol/L 3.03b ± 0.60 2.70a ± 0.75 2.81a ± 0.83 2.81a ± 0.76 0.08 0.003
LDL:HDL ratio 2.85 ± 0.90 2.66 ± 1.02 2.74 ± 0.96 2.77 ± 1.09 0.09 0.256
Triglycerides, mmol/L 1.50c ± 0.57 0.98a ± 0.30 1.18b ± 0.44 1.14b ± 0.46 0.07 <0.001
Total cholesterol:HDL cholesterol ratio 4.62b ± 1.37 4.34a ± 1.31 4.52ab ± 1.35 4.48a,b ± 1.60 0.13 0.240
NEFAs, mmol/L 0.60a,b ± 0.22 0.68b ± 0.24 0.56a ± 0.24 0.56a ± 0.15 0.04 0.015
1
Values are means ± SDs, n = 18 (1 volunteer did not complete these measurements). Means in the same row not sharing a superscript letter differ (P < 0.05). C-WM, control
weight-maintenance diet; M, maintenance diet; NEFA, nonesterified fatty acid; RS-WM, resistant-starch weight-maintenance diet; SED, standard error of the difference
between the means; WL, weight-loss diet. Data were normally distributed.
2
Analyzed by ANOVA.

1864 Johnstone et al.

TABLE 5 Total and incremental AUCs of glucose and gut hormones in plasma after consumption of M, WL, C-WM, and RS-WM
diets (measured in overweight adults over 3 h on study days 8, 29, 40, and 50)1

Variable M WL C-WM RS-WM SED P2

Glucose tAUC, mmol/L min 1140 ± 319 1100 ± 176 1100 ± 259 1050 ± 211 44 0.345
Glucose iAUC, mmol/L min 157 ± 232 106 ± 86 126 ± 180 116 ± 132 36 0.541
Ghrelin tAUC, pg/L min 16.6a ± 12.3 22.6b ± 16.8 17.0a ± 11.4 17.5a ± 10.8 1.35 <0.001
Ghrelin iAUC, pg/L min 0.13 ± 0.58 0.39 ± 0.55 0.17 ± 0.45 0.31 ± 0.83 0.11 0.086
GIP tAUC, pg/L min 180c ± 53.5 120a ± 34.2 155b ± 54.5 143b ± 57.9 11.1 <0.001
GIP iAUC, pg/L min 159c ± 49.1 106a ± 34.6 138bc ± 53.1 127a,b ± 55.4 10.5 <0.001
Insulin tAUC, pg/L min 938c ± 410 506a ± 249 665b ± 377 549a,b ± 268 76.6 <0.001
Insulin iAUC, pg/L min 782b ± 384 396a ± 211 539b ± 350 426a,b ± 223 68.6 <0.001
GLP-1 tAUC, pg/L min 7.44 ± 3.78 6.99 ± 4.48 7.44 ± 4.35 8.30 ± 5.86 0.72 0.344
GLP-1 iAUC, pg/L min 4.67 ± 2.67 4.72 ± 3.75 5.00 ± 3.82 6.05 ± 5.45 0.84 0.343
PYY tAUC, pg/L min 24.6 ± 7.07 22.5 ± 6.49 26.6 ± 8.75 26.8 ± 11.4 1.81 0.087
PYY iAUC, pg/L min 10.5 ± 6.34 9.55 ± 7.88 11.9 ± 9.27 13.7 ± 12.7 2.03 0.222
1
Values are means ± SDs, n = 18 (1 volunteer did not complete this measurement). Means in the same row not sharing a superscript letter differ (P < 0.05). C-WM, control
weight-maintenance diet; GIP, glucose-dependent insulinotropic polypeptide; GLP-1, glucagon-like peptide 1; iAUC, incremental AUC; M, maintenance diet; PYY, pancreatic
peptide YY; RS-WM, resistant-starch weight-maintenance diet; SED, standard error of the difference between the means; tAUC, total AUC; WL, weight-loss diet. Data were
normally distributed.
2
Analyzed by ANOVA.

Ruminococcus, and Faecalibacterium (Figure 4B). Genus-level compared with after the C-WM dietary period (Figure 4B).
analysis revealed a substantial increase in both Ruminococcus Similarly to the enrichment of Roseburia and Ruminococcus,
and Roseburia after RS-WM compared with the C-WM diet the increase in R. bromii was only significant when compared
period, increasing by a mean of 2.2-fold and 2.4-fold to with the WL diet period (q = 0.05; Metastats statistics)
13.4% (±8.7) and 6.4% (±5.6) relative abundance, respectively and the C-WM diet period (q = 0.15; Metastats statistics),
(Figure 4B). Although this enrichment was not significant, and there was no significant difference in mean abundance
the increase in Roseburia (q = 0.11; Metastats statistics, between the WL and C-WM diet periods. In certain individuals,
see Supplemental Methods) and Ruminococcus (q = 0.26; OTU1972 was preferentially enriched during the RS-WM
Metastats statistics) was observed in 15 of 19 and 13 of 19 diet (Figure 4C), increasing in 70% of subjects by a mean
subjects, respectively. The consistent high number of subjects of 2.8-fold (Figure 4B). This R. bromii OTU represented a
exhibiting an increase in these genera concurs with findings substantial proportion of the total fecal microbiota after the
in other studies (18), suggesting that these are important RS-WM diet, accounting for ≤30% of the total microbial
RS-responsive groups of bacteria. Interestingly, a significant population.
enrichment of Ruminococcus was observed after the RS-WM An individual OTU (OTU2064) with a 100% sequence
diet period when compared with after the WL diet period identity to Eubacterium rectale and a 99.60% sequence identity
(q = 0.005; Metastats), which was not apparent when WL was to Roseburia faecis was significantly higher after the RS-WM
compared with the C-WM diet period. diet period compared with the C-WM diet period, increasing
The enrichment of Ruminococcus after the RS-WM diet in 16 out of 19 subjects by a mean of 15.9-fold (q = 0.03;
period was attributable to an increase in the species Ruminococ- Metastats statistics; Figure 4B). In addition to the selective
cus bromii, which increased by a mean of 2.6-fold in 88% of enrichment of certain taxa, the RS-WM diet period also
subjects to a mean of 9.8% ± 8% relative abundance when triggered a significant reduction in the dominant gut genus

FIGURE 3 doneFecal SCFAs (A) and BCFAs (B) of overweight adults after consumption of study M, WL, C-WM, and RS-WM diets. Values
are means ± SEs, n = 19. Determined by capillary GC, group means are presented as percentage of total. Means not sharing a superscript
letter differ (P < 0.05). BCFA, branched-chain fatty acid; C-WM, control weight-maintenance diet; M, maintenance diet; RS-WM, resistant-starch
weight-maintenance diet; WL, weight-loss diet.

Impact of complex carbohydrates on gut microbiota 1865

FIGURE 4 Diet-induced changes in the fecal microbiota of overweight adults after consumption of study M, WL, C-WM, and RS-WM diets;
n = 19. (A) NMDS ordination of the fecal microbiota after the WM dietary periods using Bray-Curtis dissimilarity. (B) Scatterbox plots of OTUs and
summarized taxonomic groups exhibiting a substantial shift in abundance after the WM dietary periods. (C) Line plot representing the change
in relative abundance of all OTUs classified as Ruminococcus bromii after all 4 dietary periods. C-WM, control weight-maintenance diet; M,
maintenance diet; NMDS, nonmetric multidimensional scaling; OTU, operational taxonomic unit; RS-WM, resistant-starch weight-maintenance
diet; WM, weight-maintenance; WL, weight-loss diet.

Blautia when compared with the WL diet (q = 0.004; Metastats of Faecalibacterium prausnitzii after the RS-WM diet period
statistics), which was not observed between the C-WM and WL was observed when estimated by qPCR. F. prausnitzii is one
diet periods (Figure 4B). of the most abundant bacterial species in the healthy human
Diet-induced microbial changes were also examined using large intestine and its abundance is often diminished in certain
qPCR by targeting specific bacterial groups, which was in excel- disease conditions. It is considered to be a major contributor
lent agreement with the high-throughput sequencing results and to butyrate formation in the colon, which is the major energy
confirmed diet-driven increases in R. bromii and related species source for colonocytes, and also possesses anti-inflammatory
(Figure 5). A small significant increase (P = 0.042) in percentage activities (30).

1866 Johnstone et al.

FIGURE 5 Percentage of total bacteria in fecal samples from overweight adults, after consumption of study M, WL, C-WM, and RS-WM
diets, measured by qPCR; n = 19. Means not sharing a superscript letter differ (P < 0.05). C-WM, control weight-maintenance diet; F. prau,
Faecalibacterium prausnitzii; M, maintenance diet; RS-WM, resistant-starch weight-maintenance diet; WL, weight-loss diet.

Spearman rank correlation was utilized to investigate associ- biomarkers for obesity. Consistent with this, study participants
ations between genus-level microbiota abundance and multiple showed a 2-fold increase in IL-6 concentrations compared with
bodily parameters including blood glucose (millimoles per liter), the normal range (<5 pg/mL), but these differences were not
weight (kilograms), BMI (kg/m2 ), HOMA-IR, and HOMA- statistically significant (Supplemental Table 6). IL-10 concen-
β cell (Supplemental Table 6). Of the few genera showing trations were found to be extremely low and frequently below
significant correlations with various bodily parameters, the the minimum level of detection (<3.3 pg/mL), likely because the
majority exhibited very weak coefficients (Spearman ρ < 0.3), volunteers were obese and have low-grade and chronic inflam-
suggesting weak associations. The strongest correlation was mation, which decreases anti-inflammatory interleukins such
between the genus Anerostipes and blood glucose concentration as IL-10.
(P < 0.001, Spearman ρ = 0.46), revealing a moderate
positive association. It is possible that the high level of
interpersonal variation observed in this group of volunteers may
be masking other stronger microbial associations. Also, due Discussion
to the compositional nature of microbiota data, results of the It is well established that diet is a major factor shaping the
correlation analysis should be interpreted with caution. composition of the intestinal microbiota and that changes
in diet can induce specific alterations within the microbial
Satiety, digestive discomfort community (18, 19, 32–36). Consistent with this, the present
Based on AUC, ghrelin increased significantly after the WL study demonstrates that RS3 enhances a distinct group of
diet period relative to the other 3 diet periods (Table 5). GIP species within the intestinal microbiota. The most striking RS-
decreased after the WL diet period relative to M, and increased induced microbial changes within the fecal microbiota were
after both WM diet periods. There was no evidence of digestive those involving the Ruminococcaceae, primarily due to the
discomfort associated with any of the 4 dietary periods based enrichment of R. bromii in subjects. R. bromii is a known
on questionnaires (data not shown). degrader of RS (37) that has been shown previously to be
strongly promoted by Novelose RS3 in human dietary studies
Impact of diet on inflammatory biomarkers (18). Our findings strengthen the evidence that R. bromii is one
Results from inflammatory measurements (Supplemental Table of the major taxa involved in RS fermentation in the human
7) indicate that, within the same biomarker, there were no colon (18, 32, 37, 38). Consumption of the RS-WM diet also
statistically significant differences (P < 0.05) between the significantly increased representation of an OTU classified as
different diets of the study. TNF-α and some interleukins, such Eubacterium rectale, by a mean of 15.9-fold in 84% of subjects.
as IL-1β, IL-8, and IL-12p70, are classified as proinflammatory; This finding is supported by other reports (18, 32), suggesting
and in the study subjects, IL-1β was detected at concentrations a significant increase in E. rectale following consumption
within the normal range, although near the upper limit by the of RS3.
end of the study period (0–5 pg/mL). The lack of any significant change in the concentrations
Cytokine IL-6 is considered both pro- and anti-inflammatory of fecal SCFAs across the 4 diets probably reflects the fact
and in obesity, a chronic inflammatory condition, it is that the change in ND carbohydrate content between the
considered proinflammatory (31). This cytokine as well as diets was relatively small. In previous studies we reported
IL-8 are considered to be the most important inflammatory that fecal butyrate proportion and butyrate-producing bacterial

Impact of complex carbohydrates on gut microbiota 1867

populations were significantly decreased with low-carbohydrate complex interaction of psychological and physiological systems
WL diets (39). These studies, however, involved greater involving the link between the gut and brain (43). In the
differences in total carbohydrate and fiber intake compared current study, the energy density of the diets was held constant,
with the present study in which carbohydrate intakes for all an important feature in modulating short-term appetite. It is
diets were between 40% and 55% of calories. Significant generally accepted that diet composition strongly affects ad
changes were observed, however, in the BCFAs iso-butyrate libitum energy intake under laboratory (44) and free-living
and iso-valerate. Since these acids are products of branched- conditions (45), with protein highlighted as the most satiating
chain amino acid fermentation, their increase with the WL diet macronutrient (46) independent of energy density. It was also
(30% protein) compared with the initial maintenance diet (15% notable that subjects reported a similar motivation to consume
protein) is consistent with more protein being fermented in the 2 maintenance diets. Interestingly, we did detect an increase
the colon relative to carbohydrate on the WL diet, as noted in ghrelin concentration during the dieting period, suggesting
previously with WL diets that contained an increased ratio subjects were able to detect a calorie deficit.
of protein to fermentable carbohydrate and fiber (3). On the Certain soluble fibers form a viscous gel matrix in the gut,
other hand, the percentage of BCFAs did not simply change believed to slow gastric emptying and lead to a greater feeling of
with the percentage of dietary protein since BCFA values were fullness (47, 48). Viscous fibers can slow absorption of glucose
significantly lower for the RS-WM than for the C-WM diet in the small intestine and lead to lower postprandial glycemic
(although both contained 20% protein). This implies that there and insulinemic responses (49, 50). Both of these mechanisms
were higher rates of ND carbohydrate fermentation relative to are postulated to increase satiety. Insoluble fiber has limited
protein fermentation on the RS-WM diet compared with the C- effects on gastric emptying and absorption in the small intestine,
WM diet. but it may be partially fermented in the large intestine. Research
Diets containing 20% protein and 50% total carbohydrate on RS and satiety is inconsistent (51–53). de Roos et al.
(both the C-WM and RS-WM diets) were equally effective (52) reported that 30 g/d supplementation with high-amylose
in maintaining WL in overweight volunteers through a 21- corn starch over 4 wk reduced appetite relative to glucose,
d period with a high protein (30%) diet with limited caloric although subjects still maintained the cyclic hunger pattern, but
intake. These diets were also successful in maintaining the paradoxically felt less full. Raben et al. (51) also examined the
improvements in SBP and concentrations of triglycerides, total effect of RS on subjective satiety and palatability ratings in
cholesterol, and LDL. The RS-WM diet was significantly more healthy males. The test meals consisted of pregelatinized starch
successful than the C-WM diet in maintaining normal blood with 0% RS (S) or raw potato starch with 54% RS (R) together
glucose concentrations, which is in line with previous studies with artificially sweetened syrup. Scores for satiety and fullness
reporting improvements in blood glucose concentration with were significantly lower after the R meal than S meal. Thus, the
RS-supplemented diets (9, 10). replacement of digestible starch with RS results in significant
The role of RS as a fermentable carbohydrate has been reductions in subjective sensations of satiety.
examined by Robertson et al. (9, 10) in a short-term study (24- One study demonstrated that satiety-influencing gut hor-
h supplementation with 60 g RS) and a 4-wk supplementation mones were increased after rats were fed a high-RS diet for
with 30 g/d additional RS. In both studies there was a reduction 1 mo (54). It is possible that certain types or amounts of RS
in postprandial glucose and insulin concentrations and an could improve satiety by increasing concentrations of satiety
increase in systemic propionate concentrations following high hormones, GLP-1 or PYY. RS may also mediate satiety by
RS intakes. The propionate AUC values following the 4-wk altering colonic fermentation and gastric-emptying rate (53),
supplementation (10) were significantly higher (P = 0.012). This with colonic fermentation (as measured by breath hydrogen)
suggests that changes in propionate supply, a product of the positively correlated with satiety and inversely correlated with
fermentation of RS within the large bowel, may be an important gastric emptying. In the current study, the RS product used as
factor in regulation of endogenous glucose production (EGP). a fermentable carbohydrate may have influenced gastric and
Propionate is gluconeogenic and can stimulate glycolysis (40), intestinal motility and increased SCFA production, which can
but direct effects on EGP in humans have not been proven (41). modify the release of gut hormones involved in the control of
Propionate is sensed by the intestinal SCFA sensor, free fatty appetite. However, the AUC for GLP-1 remained similar after
acid receptor 2 , leading to secretion of GLP-1, which stimulates RS consumption, perhaps due to insufficient power to reach
insulin secretion. In contrast to data reported by Robinson et statistical significance. Delray et al. (55) examined short-term
al. (9, 10), Laurent et al. (41) have demonstrated that there effects of types of fiber (22 g) on appetite, using gum (soluble
were no changes in blood glucose concentration or hepatic fiber) and wheat bran (insoluble fiber) consumed at breakfast,
glucose production in response to gastric infusion of acetate and suggesting that wheat bran suppressed ad libitum intake at
propionate. In our study, the large interpersonal variation may lunchtime, with a trend towards an effect on longer-term (9–
explain why changes in the glucose AUC for RS did not reach 13 h) hunger later in the day. Pasman et al. (56) studied the
significance compared with the control. effect of 1-wk supplementation with a water-soluble fiber (guar
Moreover, structural differences and different sources of gum) in an energy-deficit diet (−2 MJ/d) and a normal energy
RS may vary according to proportion of ND starch, dietary diet and reported no change in hunger scores in the “normal”
fiber content, starch granule size, and other physicochemical diet but a reduction in the energy-deficit diet, suggesting that
differences that might result in different levels of digestibility fiber may be effective in facilitating compliance to low energy
and fermentation patterns. As a consequence, different forms intake. Other WL studies with fiber supplementation have
of RS might vary considerably in their effects on colonic reached similar conclusions (57, 58). Increasing fiber intake
microbiota and release of SCFAs. during WM, however, has limited impact on body-weight
In this study, subjects did not report a change in hunger when control (59).
placed on a high-protein WL diet. Weigle et al. (42) also reported The present study therefore has a number of strengths,
that subjects showed no increase in hunger when fed an ad which include complete diet provision for the duration and a
libitum high-protein (30% of energy) diet. Hunger involves a within-subject design to facilitate both the subjective assessment
1868 Johnstone et al.
of appetite and detection of changes in microbiota profile. 7. Morrison DJ, Preston T. Formation of short chain fatty acids by the
However, as with any dietary intervention study, it has gut microbiota and their impact on human metabolism. Gut Microbes
2016;7:189–200.
limitations, such as limited study duration, the amount and
8. Psichas A, Reimann F, Gribble FM. Gut chemosensing mechanisms. J
type of RS, and many phenotypic effects that have not been
Clin Invest 2015;125:908–17.
investigated (e.g., adiposity, age, and gender). Although 10 d is
9. Robertson DM, Currie JM, Morgan LM, Jewell DP, Frayn KN. Prior
a fairly short time frame, it is evident, from our previous studies, short-term consumption of resistant starch enhances post prandial
that changes in the gut microbiota occur rapidly following diet insulin sensitivity in healthy subjects. Diabetologia 2003;46:659–65.
shifts (18). In our statistical analysis we presented unadjusted 10. Robertson DM, Bickerton AS, Dennis AL, Vidal H, Frayn KN. Insulin-
P values for comparing treatment groups in tests of several sensitizing effects of dietary resistant starch and effects on skeletal
variables, to preserve the power of the study, so although muscle and adipose tissue. Am J Clin Nutr 2005;82:559–67.
there are clear patterns of significant differences, there is a risk 11. Robertson DM. Metabolic cross talk between the colon and
the periphery: implications for insulin sensitivity. Proc Nut Soc
that a small proportion of these are type I errors. Within our 2007;82:559–67.
recruitment we recruited people motivated to respond to a diet 12. Lockyer S, Nugent AP. Health effects of resistant starch. Nutr Bull
trial, so not a truly random sample; furthermore, we do not 2017;42(1):10–41.
present any data on the 5 dropouts. A much larger sample size 13. Li YO, Komarek AR. Dietary fibre basics: Health, nutrition, analysis,
may allow the statistical results to be generalized to a larger and applications. Food quality and safety 2017;1(1):47–59.
population or phenotype. 14. Ma Z, Hu X, Boye JI. Research advances on the formation mechanism
In conclusion, we have shown that RS3 starch supplemen- of resistant starch type III: a review. Crit Rev Food Sci 2018;24:
1–22.
tation results in specific changes in the fecal microbiota at the
15. Nicholson JK, Holmes E, Kinross J, Burcelin R, Gibson G, Jia
level of individual “diet-responsive” bacterial species. Moreover,
W, Letterson S. Host-gut microbiota metabolic interactions. Science
many of the metabolic benefits (decreased blood pressure, 2012;336:1262–7.
cholesterol) brought about by achieving WL on a high (30%) 16. Anderson PD, Mehta NN, Wolfe ML, Hinkle CC, Pruscino L, Comiskey
protein diet, can be maintained through a subsequent WM LL, Tabita-Martinez J, Sellers KF, Rickels MR, Ahima RS, et al. Innate
diet that is lower in protein and higher in total carbohydrate. immunity modulates adipokines in humans. J Clin Endocrin Metab
Notably, the consumption of RS compared with the control diet 2007; 92:2272–9.
resulted in a lower fasting glucose concentration and reduced 17. Maslowski KM, Vieira AT, Ng A, Kranich J, Sierro F, Yu D, Schilter
HC, Rolph MS, Mackay F, Artis D, et al. Regulation of inflammatory
protein fermentation. responses by gut microbiota and chemoattractant receptor GPR43.
Nature 2009;461:1282–8.
Acknowledgments
18. Walker AW, Ince J, Duncan SH, Webster LM, Holtrop G, Ze X, Brown
We thank Petra Louis for qPCR advice. The authors’ re- D, Stares MD, Scott P, Bergerat A, et al. Dominant and diet responsive
sponsibilities were as follows—AMJ, RR-G, HJF, SWG, and groups of bacteria within the human microbiota. ISME J 2011;5:220–
SHD: designed and carried out dietary trial studies, with 30.
AMJ, RR-G, HM, HJF, SWG, SHD, EN, RL-N, GR-B, CF-S, 19. David LA, Maurice CF, Carmody RN, Gootenberg DB, Button JE, Wolfe
J Harrold, J Halford, and CF conducting the related research; BE, Ling AV, Devlin AS, Varma Y, Fischbach MA, et al. Diet rapidly and
DB and AB: provided study design support and product reproducibly alters the human gut microbiome. Nature 2014;505:559–
63.
development; CF-S, XZ, JK, SR, and SS-B: carried out the work
20. Bindels LB, Walter J, Ramer-Tait AE. Resistant starches for the
on gut microbiota and associated bioinformatic analysis; EN management of metabolic diseases. Curr Opin Clin Nutr Metab Care
conducted the gut hormone analysis; GH: carried out statistical 2015;18(6):559–65.
analysis of anthropometric parameters; AMJ, HJF, SHD, and 21. Henderson L, Gregory J, Irving K, Swan G. National Diet and
SS-B: wrote the manuscript; AMJ, SHD, CF, HJF, SS-B, DB, AB, Nutrition Survey: adults aged 19 to 64 years. Volume 2: energy, protein,
and JK: had primary resp2onsibility for final content; and all carbohydrate, fat and alcohol intake. London: TSO; 2002.
authors: read and approved the final manuscript. 22. Holland B, Welch AA, Unwin ID, Buss DH, Paul AA, Southgate DAT;
Food Standards Agency. McCance and Widdowson’s the composition
of foods. 5th ed. Ministry of Agriculture, Fisheries and Food and Royal
References Society of Chemistry. London: HMSO; 1991.
23. Johnstone AM, Murison SD, Duncan JS, Rance KA, Speakman JR.
1. Johnstone AM, Horgan GW, Murison SD, Bremner DM, Lobley GE. Factors influencing variation in basal metabolic rate include fat-free
Effects of a high-protein ketogenic diet on hunger, appetite, and weight mass, fat mass, age, and circulating thyroxine but not sex, circulating
loss in obese men feeding ad libitum. Am J Clin Nutr 2008;87: leptin, or triiodothyronine. Am J Clin Nutr 2005;82(5):941–8.
44–55. 24. Flint A, Raben A, Blundell JE, Astrup A. Reproducibility, power and
2. Neacsu M, Fyfe C, Horgan G, Johnstone AM. Appetite control and validity of visual analogue scales in assessment of appetite sensations in
biomarkers of satiety with vegetarian (soy) and meat-based high protein single test meal studies. Int J Obes 2000;24(1):38–48.
diets for weight loss in obese men: a randomized crossover trial. Am J 25. Peruhype-Magalhães V, Martins-Filho OA, Prata A, Silva Lde A,
Clin Nutr 2014;100:548–58. Rabello A, Teixeira-Carvalho A, Figueiredo RM, Guimarães-Carvalho
3. Russell WR, Gratz SW, Duncan SH, Holtrop G, Ince J, Scobbie SF, Ferrari TC, Van Weyenbergh J, et al. Mixed inflammatory/regulatory
L, Duncan G, Johnstone AM, Lobley GE, Wallace RJ, et al. High- cytokine profile marked by simultaneous raise of interferon-gamma and
protein, reduced carbohydrate weight-loss diets promote metabolite interleukin-10 and low frequency of tumour necrosis factor-alpha(+)
profiles likely to be detrimental to colonic health. Am J Clin Nutr monocytes are hallmarks of active human visceral Leishmaniasis due
2011;93:1062–72. to Leishmania chagasi infection. Clin Exp Immunol 2006;146(1):
4. Ritzhaupt A, Ellis A, Hosie KB, Shirazi-Beechey SP. The characterization 124–32.
of butyrate transport across pig and human colonic luminal membrane. 26. Richardson AJ, Calder AG, Stewart CS, Smith A. Simultaneous
J Physiol 1998;15:819–30. determination of volatile and non-volatile acidic fermentation products
5. Flint HJ, Duncan SH, Scott KP, Louis P. Interactions and competition of anaerobes by capillary gas chromatography. Lett Appl Microbiol
within the microbial community of the human colon: Links between 1989;9:5–8.
diet and health. Environ Microbiol 2007;9:1101–11. 27. Storey D, Lee A, Bornet F, Brouns F. Gastrointestinal responses
6. Roberfroid M, Gibson GR, Delzenne N. The biochemistry of following acute and medium term intake of retrograded resistant
oligofructose, a non-digestible fiber: an approach to calculate its caloric maltodextrins, classified as type 3 resistant starch. Eur J Clin Nutr
value. Nutr Rev 1993;51:137–46. 2007;61(11):1262–70.

Impact of complex carbohydrates on gut microbiota 1869

28. Mariat D, Firmesse O, Levenez F, Guimarăes V D, Sokol H, Doré J, 43. Badman MK, Flier JS. The gut and energy balance: visceral allies in the
Corthier G, Furet JP. The Firmicutes/Bacteroidetes ratio of the human obesity wars. Science 2005;307(5717):1909–14.
microbiota changes with age. BMC Microbiol 2009;9:123. 44. Poppitt SD, McCormack D, Buffenstein R. Short-term effects of
29. Martınez I, Kim J, Duffy PR, Schlegel VL, Walter J. Resistant starches macronutrient pre-loads on appetite and energy intake in lean women.
types 2 and 4 have differential effects on the composition of the fecal Physiol Behav 1998;64(3):279–85.
microbiota in human subjects. PLoS One 2010;5:e15046. 45. de Castro JM. Macronutrient and dietary energy density influences on
30. Lopez-Siles M, Martinez-Medina M, Surís-Valls R, Aldeguer X, Sabat- the intake of free-living humans. Appetite 2006;46(1):1–5.
Mir M, Duncan SH, Flint HJ, Garcia-Gil LJ. Changes in the abundance 46. Halton TL, Hu FB. The effects of high protein diets on thermogenesis,
of Faecalibacterium prausnitzii phylogroups I and II in the intestinal satiety and weight loss: a critical review. J Am Coll Nutr
mucosa of inflammatory bowel disease and patients with colorectal 2004;23(5):373–85.
cancer. Inflamm Bowel Dis 2016;22(1):28–41.
47. Howarth NC, Saltzman E, Roberts SB. Dietary fiber and weight
31. Curti MLR, Jacob P, Borges MC, Rogero MM, Ferreira SRG. regulation. Nutr Rev 2001;59(5):129–39.
Studies of gene variants related to inflammation, oxidative stress,
48. Hoad CL, Rayment P, Spiller RC, Marciani L, Alonso Bde C, Traynor
dyslipidemia, and obesity: Implications for a nutrigenetic approach. J
C, Mela DJ, Peters HP, Gowland PA. In vivo imaging of intragastric
Obes 2011;2011:31.
gelation and its effect on satiety in humans. J Nutr 2004;134(9):2293–
32. Salonen A, Lahti L, Salojärvi J, Holtrop G, Korpela K, Duncan SH, 300.
Date P, Farquharson F, Johnstone AM, Lobley GE, et al. Impact of
49. Behall KM, Scholfield DJ, Hallfrisch JG, Liljeberg-Elmståhl HG.
diet and individual variation on intestinal microbiota composition and
Consumption of both resistant starch and beta-glucan improves
fermentation products in obese men. ISME J 2014;8:2218–30.
postprandial plasma glucose and insulin in women. Diabetes Care
33. Magoc T, Salzberg SL. FLASH: fast length adjustment of short reads to 2006;29(5):976–81.
improve genome assemblies. Bioinformatics 2011;27:2957–63. 50. Flammang AM, Kendall DM, Baumgartner CJ, Slagle TD, Choe YS.
34. Daly K, Darby AC, Hall N, Wilkinson MC, Pongchaikul P, Bravo D, Effect of a viscous fiber bar on postprandial glycemia in subjects with
Shirazi-Beechey SP. Bacterial sensing underlies artificial sweetener- type 2 diabetes. J Am Coll Nutr 2006;25(5):409–14.
induced growth of Lactobacillus. Environ Microbiol 2016;18: 51. Raben A, Tagliabue A, Christensen NJ, Madsen J, Holst JJ, Astrup
2159–71. A. Resistant starch: the effect on postprandial glycemia, hormonal
35. Kelly J, Daly K, Moran AW, Ryan S, Bravo S, Shirazi-Beechey SP. response, and satiety. Am J Clin Nutr 1994;60(4):544–51.
Composition and diversity of mucosa-associated microbiota along the 52. de Roos N, Heijnen ML, de Graaf C, Woestenenk G, Hobbel E. Resistant
entire length of the pig gastrointestinal tract; dietary influences. Environ starch has little effect on appetite, food intake and insulin secretion of
Microbiol 2017;19(4):1425–38. healthy young men. Eur J Clin Nutr 1995;49(7):532–41.
36. Franks AH, Harmsen HJM, Raangs CG, Jansen GJ, Schut F, Welling 53. Nilsson AC, Ostman EM, Holst JJ, Björck IM. Including indigestible
GW. Variations of bacterial populations in human feces quantified carbohydrates in the evening meal of healthy subjects improves glucose
by fluorescent in situ hybridization with group specific 16S rRNA tolerance, lowers inflammatory markers, and increases satiety after a
targeted oligonucleotide probes. Appl Environ Microbiol 1998;64: subsequent standardized breakfast. J Nutr 2008;138(4):732–9.
3336–45. 54. Keenan MJ, Zhou J, McCutcheon KL, Raggio AM, Bateman HG, Todd
37. Ze X, Duncan SH, Louis P, Flint HJ. Ruminococcus bromii is a keystone E, Jones CK, Tulley RT, Melton S, Martin RJ, et al. Effects of resistant
species for the degradation of resistant starch in the human colon. ISME starch, a non-digestible fermentable fiber, on reducing body fat. Obesity
J 2012;6(8):1535–43. (Silver Spring) 2006;14(9):1523–34.
38. Edgar RC. Search and clustering orders of magnitude faster than BLAST. 55. Delray HJ, O’Sullivan KR, Fletcher RJ, Blundell JE. Effects of amount
Bioinform 2010;26:2460–1. and type of dietary fibre (soluble and insoluble) on short-term control
39. Duncan SH, Lobley GE, Holtrop G, Ince J, Johnstone AM, Louis P, Flint of appetite. Int J Food Sci Nutr 1997;48(1):67–77.
HJ. Human colonic microbiota associated with diet, obesity and weight 56. Pasman WJ, Saris WH, Wauters MA, Westerterp-Plantenga MS. Effect
loss. Int J Obes (Lond) 2008;32(11):1720–4. of one week of fibre supplementation on hunger and satiety ratings and
40. Venter CS, Vorster HH, Cummings JH. Effects of dietary propionate energy intake. Appetite 1997;29(1):77–87.
on carbohydrate and lipid metabolism in healthy volunteers. Am J 57. Astrup A, Vrist E, Quaade F. Dietary fibre added to very low calorie diet
Gastroenterol 1990;85(5):549–53. reduces hunger and alleviates constipation. Int J Obes 1990;14(2):105–
41. Laurent C, Simoneau C, Marks L, Braschi S, Champ M, Charbonnel B, 12.
Krempf M. Effect of acetate and propionate on fasting hepatic glucose 58. Rigaud D, Ryttig KR, Leeds AR, Bard D, Apfelbaum M. Effects of
production in humans. Eur J Clin Nutr 1995;49(7):484–91. a moderate dietary fibre supplement on hunger rating, energy input
42. Weigle DS, Breen PA, Matthys CC, Callahan HS, Meeuws KE, and faecal energy output in young, healthy volunteers: a randomized,
Burden VR, Purnell JQ. A high-protein diet induces sustained double-blind, cross-over trial. Int J Obes 1987;11(Suppl 1):73–8.
reductions in appetite, ad libitum caloric intake, and body weight 59. Aston LM, Stokes CS, Jebb SA. No effect of a diet with a reduced
despite compensatory changes in diurnal plasma leptin and ghrelin glycaemic index on satiety, energy intake and body weight in overweight
concentrations. Am J Clin Nutr 2005;82(1):41–8. and obese women. Int J Obes (Lond) 2008;32(1):160–5.

1870 Johnstone et al.