Impact of dietary protein quantity on the non-dysbiotic human microbiome: a controlled feeding study

Abstract

The gut microbiome plays an essential role in human health, and alterations in its composition have been associated with a range of gastrointestinal and metabolic disorders. Diet is one of the most important modifiable factors influencing the gut microbiome, with dietary protein known to affect microbial composition and metabolic activity. However, the impact of varying levels of dietary protein on the human gut microbiome remains incompletely understood. This controlled crossover feeding study examined whether diets providing 10% versus 25% of total energy intake from protein alter stool microbiome profiles in healthy adults. Ten participants, four men and six women aged 22–34 years, completed a controlled crossover feeding trial provisioned by the Bionutrition Unit of the Clinical & Translational Research Center at Oregon Health & Science University. Each diet phase consisted of a 3-day run in diet (18% protein) followed by either a 10% protein (lower-protein) or a 25% protein (higher-protein) diet for 7 days in a randomized order, and then, after a washout period, the run in diet followed by the alternate dietary intervention. Anthropometric measurements, dietary tolerance, diet satisfaction, and stool samples for microbiome 16 S rRNA gene sequencing and short chain fatty acids (SCFA), analysis were collected at 5 time points throughout the study. Body mass index was lower after than before the higher-protein diet (p < 0.01). Minor dietary tolerance symptoms included constipation in one participant on the higher-protein diet, mild discomfort in three participants, and fatigue in seven participants on the lower-protein diet. Diet satisfaction was similar (p = 0.28). Stool microbiome composition varied more by individual (57.8%) than by diet (3.5%). No significant changes in dysbiosis scores (p = 0.68), microbiome diversity (p > 0.05), or SCFA concentrations (p > 0.05) were observed. Short-term consumption of dietary protein at 10% or 25% of total energy intake did not result in significant differences in microbiome diversity or SCFA concentrations. This observation may be due in part to the short duration of the study intervention, the subjects’ identity factors, and the different sources of protein in the two intervention diets.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-026-46663-y.

Introduction

Diet is a major environmental determinant of gut microbiome composition and function, interacting with host genetics and nutrient availability to shape microbial ecology and metabolic activity^1,2. Protein, as the primary dietary source of nitrogen, along with micronutrients such as iron and zinc, represents an important limiting resource for gut microbes and can influence microbial biomass, community structure, and metabolic pathways³. Large genome microbiome association studies further demonstrate that host genetic variation is a major determinant of gut microbial composition, particularly among ecologically important taxa, through mechanisms involving nutrient availability, host-derived substrates, and microbial metabolic specialization⁴.

Once the gut microbiome reaches maturity in early childhood, diet becomes the primary driver of compositional and functional changes^5–7. Dietary interventions can rapidly alter microbial communities, with measurable shifts in microbiome composition occurring within days of dietary change^6,8. Dietary interventions have therefore become an important focus of research aimed at understanding how nutrition can shape microbial communities and their metabolic outputs. Dietary patterns rich in plant-derived foods, such as the Mediterranean diet, have been associated with greater microbial diversity and enrichment of taxa linked to beneficial metabolic and immune effects⁹.

In contrast, the role of dietary protein in shaping the gut microbiome remains comparatively understudied in human nutrition. Animal studies suggest that both the quantity and source of dietary protein can substantially influence microbial ecology and metabolic activity within the gut. For example, high-protein diets, particularly those rich in animal-derived proteins such as casein or red meat, have been associated with reduced microbial diversity and increased abundance of certain taxa involved in protein fermentation pathways^10,11. However, data from controlled human studies remain limited, constraining our understanding of how protein intake influences microbial community structure and function. To address this gap, we conducted a controlled feeding study using a randomized crossover design to evaluate the impact of dietary protein quantity on gut microbiome composition and short-chain fatty acid (SCFA) production in healthy adults. Participants completed two 10-day dietary interventions consisting of a 3-day run-in diet¹² followed by a 7-day diet providing either lower protein (10% of total energy intake) or higher protein (25% of total energy intake), with fiber intake held constant and diets administered in random order (Fig. 1A). We hypothesized that increasing dietary protein intake would alter gut microbial community structure and metabolic activity, reflected by measurable differences in stool microbiome composition and SCFA concentrations. These findings may help clarify how macronutrient composition influences gut microbial ecology and provide insight into the role of dietary protein in shaping human microbiome function.

Fig. 1.

Study design diagram.

Results

Participant characteristics, change in weight, and dietary tolerance

Study participants included four men and six women who were between 22 and 34 years of age. All participants reported compliance with consuming only the food provided and nothing else during the two experimental dietary phases. Mean Body Mass Index (BMI) at the start of the study was 24.0 kg/m² (median 24.5 kg/m², minimum 21.4 kg/m², maximum 27.1 kg/m²) and was significantly lower after the higher protein diet (23.3 kg/m^2, p = 0.003; Table 1). This lower mean BMI after the higher protein diet is likely due to the higher protein diet as there were not significant effects of protein diet sequence or carryover effects, assessed by testing for differences between run-in diet phases, for BMI (p > 0.05 for all). There were no reports of significant dietary intolerance symptoms during the two dietary interventions, although one participant reported abdominal pain and was unable to provide a stool sample following the higher protein diet phase likely due to diet-induced constipation. Three participants reported minor gastrointestinal symptoms (looser stools, mild abdominal pain, gassiness) while consuming the lower protein diet. Seven participants noted slight to moderate fatigue while on the lower protein diet of whom two reported more challenging exercise recovery. Average stool Bristol scores were 4 after each dietary intervention indicating normal stool consistency. Subjects were satisfied with the food provided as indicated by average dietary satisfaction scores of 4.5 for the higher protein diet and 4 for the lower protein diet on a scale of 1 to 5 (p = 0.28).

Table 1.

Results of linear mixed effect model testing for the influence of diet and participant sex on Body Mass Index with participant identity as a random effect.

Predictors	Body mass index
	Estimates	p
(Intercept)	24.4 (22.6–26.3)	< 0.01
Higher protein diet	-0.4 (-0.7 – -0.2)	< 0.01
Lower protein diet	-0.2 (-0.5 – -0.0)	0.05
Run-In diet	-0.2 (-0.4–0.0)	0.07
Participant sex	-1.1 (-4.0–1.8)	0.45
Random Effects
σ2	0.08
τ00 participant	4.99
ICC	0.98
N participant	10
Observations	49
Marginal R2 / Conditional R2	0.061 / 0.986

Microbiome varies significantly among participants

Six hundred and twelve Amplicon Sequence Variants (ASVs) were detected across the 49 stool samples collected. Reads were subsampled to 14,900 reads per sample after quality control and filtering. Extraction kit and PCR controls generated at most 112 reads. These ASVs were included in the analysis as the decontam package did not identify any potential contaminant ASVs. Together, this indicates minimal, if any, contamination from sample processing and preparation of sequencing libraries. Coverage plots indicate that the sequencing depth in this study was sufficient to cover the diversity present (Supplementary Figure S4).

The microbiomes in this study were dominated either by members of the Bacteroides or Prevotella, as is typical of human microbiomes⁶. Firmicutes were also present in all samples, including Faecalibacterium (Fig. 2A).

Fig. 2.

Stool microbiome composition. (A) Relative abundance of genera of the four most dominant bacterial phyla detected in the stool samples, organized by study participant and diet phase. (B) Unconstrainted principal coordinates ordination of microbiome samples from robust Aitchison dissimilarity. (C) Partial dbRDA ordination conditioned on participant identity with dietary type as the constraining variable.

Overall, the composition of the gut microbiome varied substantially among study participants (Fig. 2A and B). To isolate potential variation in microbiome composition due to experimental diets, a partial dbRDA was run conditioned on participant identity with dietary type (lower or higher protein composition) as the constraining variable. Diet influenced microbiome composition, although the effect was relatively small (diet explained 3.5% of the variance in microbiome composition). In contrast, individual variation explained 57.8% of the variance in microbiome composition. When only the variation attributed to diet was considered, the samples from the run-in and lower-protein diets formed separate clusters in ordination space, while samples from the higher-protein diet formed an intermediate cluster with the baseline samples (Fig. 2C). There were no microbial taxa that varied consistently or significantly in relative abundance between the two dietary phases among our participants (Supplementary Figure S1). Dysbiosis scores did not differ significantly between samples collected at baseline and after the lower protein diet (𝜒² = 0.006, p = 0.94) or at baseline and after the higher protein diet (𝜒² = 0.167, p = 0.68). Microbiome composition did not differ significantly between the two standardized diet samples, indicating a lack of carryover effects of the initial dietary intervention on the subsequent dietary intervention (p > 0.05). Consuming the provided diets (run-in, lower-protein, and higher-protein) did not reduce variation in microbiome composition among participants relative to their baseline samples (p > 0.05 for permutation test of homogeneity of multivariate dispersions).

Microbiome diversity (observed richness, Shannon and inverse Simpson indices) did not differ significantly between higher- and lower-protein diets, subject sex, or their interaction (Supplementary Figure S2). The inverse Simpson index was influenced by participant identoty which explained 74% of the variation in this diversity measure. Participant identity did not significantly affect the model fits for observed richness and Shannon index. To confirm that these results were not influenced by carryover effects from the previous diet, diversity measures were compared between the two standard diet periods (i.e. before and after the first experimental diet). No significant difference was observed in the diversity of the microbiome between the two standard samples for subjects receiving the higher-protein diet or those receiving the lower-protein diet during the initial experimental phase. Variance in diversity measures did not differ significantly between baseline samples and the standardized, lower-, or higher-protein dietary interventions (p > 0.05 for all), further indicating that microbiome diversity in healthy individuals is not influenced by the levels of dietary protein tested in this study.

SCFA concentrations are not influenced by the amount of dietary protein

Stool concentrations of SCFA were not significantly influenced by the higher- or lower-protein diets, participant sex, or their interaction (Supplementary Figure S3). Participant identity explained 20–62% of the variance in SCFA concentrations and did not significantly improve model fit for any of the SCFAs. To confirm that these results were not influenced by carryover effects from the previous dietary intervention, we compared SCFA concentrations in samples from day 4 of each dietary intervention. We found no significant difference in SCFA concentrations between the two run-in diet samples for participants receiving the higher- or lower-protein diet during the initial experimental phase. Variance in SCFA concentrations did not differ significantly between baseline samples and samples collected after the run-in, lower-, or higher-protein dietary interventions (p > 0.05 for all), further indicating that SCFA concentrations in healthy individuals are not influenced by the variation in dietary protein tested in this study.

Sequencing data were deposited in the National Center for Biotechnology Information (NCBI) Sequence Read Archive under accession number PRJNA1310235.

https://www.ncbi.nlm.nih.gov/sra/?term=PRJNA1310235.

Methods

Study design

Ten participants who met inclusion criteria and provided verbal and written consent were enrolled in the study (Table 2). The study followed a randomized-order, cross-over design, allowing each participant to serve as their own control to minimize the impact of inter-individual differences on outcomes. (Fig. 1)

Table 2.

Participant eligibility criteria.

Inclusion criteria	Exclusion criteria
Age: 20–40 years	Any known or suspected intestinal disorder, including but not limited to irritable bowel syndrome, Inflammatory Bowel Disease, celiac disease, food sensitivities, food allergies, significant food restrictions
Body Mass Index: 18.9–24.9 kg/m2	Hormonal disorders
General good health	Use of chronic medications including both oral and subcutaneous contraceptives ((non-hormone secreting intrauterine device accepted)
Use of Vitamin D supplements allowed	Use of oral or intravenous antibiotics in the last three months
Willingness to abstain from alcohol consumption during the study intervention	Daily probiotic use (pill form)
Willingness to consume both the lower- and higher-protein diets	Daily multi-vitamin, except vitamin D supplements

Participants were provided with pre-weighed meals containing foods providing 100% of their estimated energy needs for weight maintenance. All study-related foods and beverages were prepared and provided by the Bionutrition Unit at the Clinical and Translational Research Center (CTRC) in the Oregon Clinical and Translational Research Institute (OCTRI). After a three-day dietary run-in period, a two-day repeating menu cycle was used for each 7-day dietary phase. During each dietary phase, participants were asked to consume only the food provided to them, to eat all the food provided to them, and to eat nothing else. Breakfast, lunch, and dinner, as well as snacks, were packaged individually, placed in a cooler, and sent with the participant to consume on the assigned day. Complete sets of meals were provided on days 1, 4, and 7 during each dietary phase.

The estimated energy requirement for weight maintenance for each individual was calculated using the Mifflin-St. Jeor equation¹³ in conjunction with the Boothby and Berkson Food Nomogram¹⁴ taking into account the participant’s sex, age, weight, height, and physical activity level. ProNutra feeding study software (Viocare Technologies, Princeton, NJ) was used to calculate the amounts of foods needed to meet the established nutrient composition of each participant’s study diets, including energy and macro- and micronutrient intakes^13,15. Diets were designed to contain varying protein sources from a typical American diet, and dietary protein was adjusted by altering the amount of dairy, eggs, poultry, beef, legumes, and fish at each meal to meet intervention protein targets.

The first 3 days of each 10-day dietary intervention were a run-in diet of moderate protein content (18% of total energy intake), followed by 7 days of either a lower (10% of total energy intake) or higher (25% of total energy intake) protein content diet. Dietary fiber intake was held constant at 25 g/2000 kcal/d across the dietary interventions. A 4-week washout period, when participants consumed their typical diet, separated the two research diet interventions. Macronutrient composition of the run-in diet and the lower and higher protein diets is shown in Table 3. Detailed diet menus are included in the Supplementary material. Subjects evaluated diet satisfaction on a scale from 1 to 5 at the end of each dietary intervention.

Table 3.

Macronutrient distribution of study diets.

Provided Diet		Energy (kcals)	Protein (grams)	Total lipid (grams)	Total carb (grams)	Total Fiber (grams)	Fiber/2000 Kcals	% Protein	% Fat	% Carb
Standard	Ave	2569.3	96.7	100.4	331.6	34.0	26.5	14.8	34.5	50.7
	SD	481.2	18.4	19.1	61.0	6.5	1.4	0.2	0.5	0.5
Low Protein	Ave	2556.3	64.9	128.5	301.4	32.1	25.0	9.9	44.1	46.0
	SD	475.6	13.0	23.7	56.9	6.7	1.8	0.2	0.7	0.6
High Protein	Ave	2559.6	161.7	105.0	254.6	33.0	25.8	24.8	36.2	39.1
	SD	470.3	29.8	20.1	45.3	6.1	1.3	0.1	0.7	0.7

Sample collection and processing

Study participants were provided with stool collection containers (Vitality Medical, Salt Lake City, UT). Samples of the first morning bowel movement (stool) from each participant were collected once before starting the dietary intervention (baseline, same for both diet phases), once on day 4 of each dietary intervention after the 3-day run-in diet, and once on day 10 at the end of each dietary intervention (5 samples total per participant)¹⁶. Stool consistency was graded utilizing Bristol stool charts at the time of collection. Samples from each stool collection were transferred into stool collection containers (Covidien Commode Stool Collector, Vitalimity Medical, Salt Lake City, UT). The remaining stool sample was homogenized, and a weighed sample was transferred to a 2-ml screw-cap Eppendorf tube and stored frozen at -80 °C for SCFA analysis. The remaining stool sample was frozen at -80 °C for microbiome analysis as illustrated in Fig. 1A. Samples were refrigerated prior to be being turned into the lab and processed within 24 h of collection.

Microbiome characterization

DNA was extracted from approximately 0.2 g of each stool sample with the QIAamp Fast DNA Stool Mini kit (Qiagen). The V4 region of 16S rRNA genes was amplified with dual-indexed primers 515F (5’ - GTGCCAGCMGCCGCGGTAA) and 806R (5’ – GGACTACHVGGGTWTCTAAT). Pooled amplicon libraries, including kit and PCR negative controls, were sequenced on a MiSeq instrument (Illumina, San Diego, CA) in a 2 × 150 bp run at the University of Oregon Genomics & Cell Characterization Core Facility.

Amplicon sequence variants (ASVs) were resolved with the DADA2 pipeline¹⁷, which removes low-quality and chimeric sequences, implemented with the QIIME 2 microbiome bioinformatics platform¹⁸. Taxonomy was assigned to ASVs with a naïve Bayes classifier pre-trained on the Silva 138 SSU database^19,20. Potential contaminants were identified with the decontam R package using the prevalence method²¹. Amplicon sequence variants that were not classified to either domain of Bacteria or Archaea, and ASVs classified as chloroplast or mitochondria sequences, were excluded from analysis.

Short chain fatty acid measurement

SCFA concentrations, including butyrate, propionate, and acetate, were measured in preprocessed stool samples by gas chromatography (GC)/mass spectroscopy (MS) in the OHSU Bioanalytical Shared Resource Pharmacokinetics Core Laboratory. Just before analysis, samples were thawed, lyophilized, and then analyzed by GC/MS as previously described²².

Statistical analysis

All statistical analyses were conducted with R (R Core Team, 2024). To test for the influence of dietary interventions on univariate measures (Body Mass Index, alpha diversity indices, and SCFA concentrations), mixed-effect models were run. Full models included diet type and participant sex as main effects with participant identity as a random effect. Dietary intervention could also influence these measures by reducing variance relative to the participants’ baseline diets. Levene’s test was used to determine whether variance among participants was reduced during the trial diets.

The difference in microbiome composition between samples, or beta diversity, was assessed with the robust Aitchison distance metric²³. This variation was visualized with principal coordinates analysis plots. The influence of diet and participant sex on microbiome composition were tested with constrained distance-based redundancy analysis (dbRDA) conditioned on participant identity. To determine whether specific microbial ASVs or genera differed in abundance between the higher- and lower-protein diets, we used the ALDEx2 analysis tool²⁴. Dysbiosis scores were calculated following the combined alpha beta diversity approach as described by Santiago et al.²⁵ Dysbiosis scores of the higher- and lower-protein diet samples were compared to the baseline samples with Kruskal-Wallis rank sum tests.

Ethical considerations

This study was approved by OHSU Institutional Review Board, Study#00019369. The study was implemented following guidelines of The Belmont Report and followed ethical principles of the Declaration of Helsinki. The study was conducted between April and December 2021 after COVID pandemic restrictions were partially lifted. Subjects were recruited using a research volunteer database, campus advertising, the OHSU research opportunities web page, and word of mouth.

Grant support

University of Oregon – Oregon Health & Science University SEED Award Medical Research Foundation, Oregon Health & Science University Faculty Innovation Award, Oregon Health & Science University.

Discussion

In this tightly controlled, randomized-order crossover feeding study, short-term variation in dietary protein intake (10% vs. 25% of total energy) did not significantly alter gut microbiome composition, diversity, dysbiosis indices, or short-chain fatty acid (SCFA) concentrations in healthy adults. Microbiome composition varied substantially between individuals but only modestly by diet, with protein intake explaining a small proportion of overall variation. Minor gastrointestinal symptoms occurred primarily during the lower-protein phase, and a modest but significant reduction in BMI was observed following the higher-protein diet, consistent with prior reports of protein-associated weight loss²⁶. Overall, these findings suggest that short-term modulation of protein intake alone is insufficient to perturb a stable, non-dysbiotic microbiome, supporting the concept of microbiome resilience in healthy hosts.

The absence of major microbiome shifts in our cohort contrasts with prior dietary intervention studies reporting rapid microbial changes within 24 h of dietary modification^8,27. This discrepancy may be explained by several factors, including the maintenance of constant fiber intake across diet phases, which likely stabilized microbial substrate availability and saccharolytic metabolism. Ensuring equivalent fiber intake required compensatory adjustments in carbohydrate and fat sources, highlighting the inherent challenges of isolating macronutrient effects in controlled feeding studies. Additionally, the relatively short intervention duration may have limited the detection of slower, community-level microbial adaptations. Notably, this study was conducted during and immediately following COVID-19 isolation measures, which may have influenced baseline microbiome composition; however, the controlled design minimized external variability and strengthened internal validity.

These findings should also be interpreted in the broader context of microbiome variability across health states. Microbial communities in healthy individuals tend to exhibit substantial ecological stability, whereas microbiomes in conditions characterized by inflammation or metabolic disturbance often display reduced diversity and greater responsiveness to environmental perturbations^28,29. Shifts in microbial community structure have been linked to altered metabolic outputs, including changes in amino acid metabolism, sulfur pathways, and short-chain fatty acid production. Constraint-based metabolic modeling studies further suggest that reduced microbial diversity can constrain metabolic capacity, leading to altered metabolite profiles and accumulation of intermediate substrates³⁰. These observations indicate that baseline microbial community structure strongly influences the magnitude of dietary effects. Dietary protein represents a relevant but comparatively understudied macronutrient in this context. Experimental models suggest that both protein quantity and source influence microbial composition and metabolic activity. Animal studies demonstrate that different protein sources can induce distinct microbial and enzymatic responses, including changes in amino acid degradation pathways and microbial energy utilization. Human studies further suggest that protein source may be more influential than total protein quantity, with higher intake of animal-derived proteins associated with enrichment of protein-fermenting taxa and lower concentrations of beneficial fermentation metabolites, while plant-derived proteins are often associated with greater microbial diversity and metabolite production^31,32. In our study, protein sources reflective of a typical Western diet were used, which may partially explain the limited microbiome changes observed in this healthy cohort.

This study’s small sample size limits generalizability, particularly across life-stage groups where microbiome composition may differ substantially. Stool microbiome analysis served as a practical proxy for the colonic microbiome, capturing luminal communities but not fully representing mucosa-associated or proximal colonic taxa, a known limitation³³. Maintaining consistent energy intake while adjusting macronutrient sources also introduced potential variability in digestion, satiety, and microbial substrate availability. Our study did not analyze products of protein fermentation in the stool, specifically ammonia, phenols, hydrogen sulphide and amines which could have shown more differences with dietary protein changes³⁴. Blinding participants to the diet phase was not feasible because dietary differences were readily apparent.

We chose to sequence the V4 region of the 16 S region rRNA gene as a cost-effective strategy to obtain high-quality reads. In most microbial species this region is approximately 254 bp and it only varies in length by a few base pairs. This makes 2 × 150 bp sequencing appropriate for this region and this approach is used in numerous microbiome studies³⁵.

Despite these limitations, this study provides insight into the relationship between dietary protein intake and gut microbiome stability in healthy adults. The findings suggest that in the absence of baseline microbial disruption, the gut microbiome may be relatively resistant to short-term changes in dietary protein quantity when other key substrates such as fiber remain constant. Future research should focus on longer term interventions, disease specific cohorts, and precision nutrition strategies that integrate host genetics, microbial ecology, disease stage and dietary composition to optimize outcomes in microbiome driven inflammatory conditions.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

ADMR

Acceptable Dietary Macronutrient Range

ASV

Amplicon sequence variants

dbRDA

Distance-based redundancy analysis

OCTRI

Oregon Clinical and Translational Research Institute

SCFA

Short Chain Fatty Acids

Author contributions

AKH-study concept, design, implementation, data processing, main manuscript preparationKA- statistics plan, data processing, microbiome analysis, manuscript preparationAH- diet design and implementation, manuscript reviewJG-diet design and implementation, manuscript reviewDS-study design, mentorship, manuscript reviewBB- microbiome part of study design, data analysis, mentorship, manuscript review.

Data availability

Sequencing data were deposited in the National Center for Biotechnology Information (NCBI) Sequence Read Archive under accession number PRJNA1310235.https://www.ncbi.nlm.nih.gov/sra/?term=PRJNA1310235.

Declarations

Competing interests

The authors declare no competing interests.