Dietary Pattern-Induced Gut Microbiota Differences Are Associated with White Matter Volume Changes in Middle-Aged Female Macaques

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The manuscript reports a randomized dietary intervention in nonhuman primates comparing Mediterranean vs Western diet patterns, linking diet to gut microbiota (notably Oscillospira), fecal SCFAs, circulating BCAAs, insulin response metrics, sCD14, and longitudinal MRI white-matter (WM) volume change. The topic is important and the dataset is potentially strong, but there are some key issues in microbiome compositional statistics, multiplicity, and methodology, my observations are listed below:

The manuscript uses mediation models, but (as currently described) mediator and outcomes are not consistently ordered in time (insulin AUC measured earlier than end-of-study microbiome). The author should provide a timeline table listing every variable with exact month/timepoint (baseline, interim, end) and sample sizes per assay.

How do you justify mediation for insulin AUC measured earlier than microbiome/SCFAs (if that is the case in your timeline)?

The authors used relative abundance (%) and parametric tests without addressing compositionality/zero-inflation which is a major statistical weakness. The authors should report if were preprocessing steps affecting compositional analysis: filtering thresholds, zero handling, normalization.

What transformations/compositional methods were used for microbiome abundances, and what diagnostics support parametric testing?

Minimum-value imputation for missing BCAA values may bias results if missingness is not random. Please report missingness per metabolite, explain why this imputation was used, and confirm results are similar without imputation.

Units and normalization for fecal SCFAs are unclear. Please specify measurement units and normalization method.

Mediation methods are not sufficiently described. Please specify the software used, model types, covariates, and bootstrap settings (including number of simulations), and add a sensitivity analysis or tone down causal claims.

Please ensure units, transformations, and sample sizes are clearly stated in legends and tables.

Provide assay-specific n and missingness (microbiome, SCFAs, BCAAs, insulin AUC, sCD14, MRI)

The conclusions may overstate causality relative to the data and analyses. Please temper causal interpretation and better acknowledge limitations (timing of measurements, multiple testing, and statistical assumptions).

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

This study, through a long-term dietary intervention experiment, systematically investigated the effects of Mediterranean and Western diets on gut microbiota, metabolites, insulin resistance, and brain white matter structure, proposing a "diet-microbiota-brain axis" regulatory mechanism. The research design is sound, the sample size is adequate, the data analysis methods are advanced, and the conclusions hold significant translational value. 

Specific Suggestions:
1. The abstract is somewhat lengthy, with some repetition of results and a lack of structural clarity. It is recommended to streamline the abstract according to a "Background/Objectives—Methods—Results—Conclusions" structure, highlighting the most critical findings (e.g., the mediating role of Oscillospira) and key statistical significances.
2. Introduction: 
2.1 While rich in background information, the description of Oscillospira is slightly verbose, with an overabundance of citations in some parts. It is recommended to streamline the description of the relationship between Oscillospira and various diseases, focusing on its potential role in metabolism and the nervous system. And clearly state the three main hypotheses of this study to enhance logical flow.
2.2 Lines 115-120: The text jumps abruptly from "IR affects white matter microstructure" to "microbial metabolites may directly affect the brain," lacking a smooth transition. It is suggested to add a transitional sentence.
3. Results: Figures 3 and 4 contain substantial information, but results are described too succinctly, lacking interpretation of trend-level findings.
4. Discussion:
4.1 The discussion is lengthy, with some paragraphs repeating content from the Introduction.
4.2 The mechanistic discussion of Oscillospira tends to be speculative, lacking direct evidence.
4.3 There is overlap in content between Figure 1 and Figure 5; It is recommended to merge Figure 1 and Figure 5 into a single mechanistic diagram highlighting key pathways.
4.4 The study used only female animals, but the discussion fails to address how this limits the generalizability of the findings.
5. Some references are relatively old. Prioritize citing literature from the last 5 years, especially reviews and clinical studies.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

1. The title can be more precise and less convoluted, e.g., "Dietary Pattern-Induced Gut Microbiota Differences Are Associated with White Matter Volume Changes in Aging Female Macaques."
2. In the introduction, probiotic interventions are presented in a largely positive light without mention of the notable study limitations and the fact that the shift in the gut microbiota may be transient as several treatment trials for probiotics have failed to find lasting effects and significant alterations in gut microbiome (citation: pubmed.ncbi.nlm.nih.gov/36986088).
3. How should the sample size be considered? The trial has 38 animals total, with “slightly varying” sample sizes across analyses.
4. Please clearly distinguish biological replicates from technical replicates.
5. With the use of vertebrates for experiments, besides stating the necessary ethics approval, suggest to include how the animals were treatment humanely or how the study tried to adhere to the 3Rs principles in animal research.
6. Animals were housed in small social groups (4–5 per enclosure); social status and stress physiology can strongly affect metabolism, immune activation, feeding behavior, and microbiome composition. Treating each monkey as fully independent in simple t-tests and Pearson correlations is questionable, unless the analysis adjusts for group/pen (and ideally dominance rank).
7. Pearson correlation assumptions (linearity, normality, outliers) are not convincingly checked; the outlier discussion for insulin is ad hoc.
8. Mediation analysis demands showing assumptions and covariate controls. The authors describe doing causal mediation when a variable is diet-affected and associated with Oscillospira, but do not describe what covariates were included, whether mediator-outcome confounding is plausible and assessed, and whether temporal precedence is satisfied (it is not clearly) etc.
9. 16S pipeline is dated and coarse for genus-level mechanistic claims. The authors use OTU clustering at 97% identity with Greengenes in QIIME but this is less accurate than modern ASV-based methods and can misclassify/merge taxa. This is problematic when the whole story hangs on “Oscillospira.”
10. Why did the authors not measure compositionality? Relative abundance can change simply because other taxa change. Without absolute quantification (e.g., qPCR, spike-ins) you cannot tell if Oscillospira truly increased or if others decreased.
11. Discussion of study limitations inadequate. Fecal SCFAs are an imperfect proxy for production/absorption/systemic exposure. They measure fecal acetate/propionate/butyrate only. White matter “volume change” is also a blunt surrogate for integrity/myelination. What about DTI/DWI metrics (FA/MD)?

Comments on the Quality of English Language

'Oscillospria abundance' - spelling error.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 4 Report

Comments and Suggestions for Authors

This manuscript presents a unique dataset derived from a long-term (31-month) randomized controlled trial in nonhuman primates (Cynomolgus macaques). The authors investigate the impact of Mediterranean versus Western diet compositions on the Oscillospira, metabolomics (SCFAs, BCAAs), systemic inflammation (sCD14), insulin resistance. The use of a primate model to bridge the gap between rodent mechanism and human observation is a significant strength. However, given the relatively modest sample size (N=38 total, split into two groups), the statistical robustness of these mediation claims is questionable. Furthermore, the analysis of correlations often risks conflating treatment effects with mechanistic associations (i.e., the correlations may be driven solely by the group separation rather than a continuous biological relationship). The absence of cognitive or functional behavioral data also limits the translational impact of the "brain aging" claims. The manuscript requires rigorous statistical re-evaluation, particularly regarding multiple comparison corrections and the validity of the mediation model, alongside a more conservative interpretation of "causality."

Title & Abstract: The phrase "Nonhuman Primate Model of Aging" is slightly misleading. The methods state the animals are ~8.8 years old (middle-aged). While this is relevant to aging trajectories, it is not a model of geriatric aging.

Abstract: The phrase "presenting causal pathways" is too strong for the statistical evidence provided. Mediation analysis on observational data within an RCT suggests statistical mediation, but without manipulating Oscillospira directly (e.g., FMT or probiotics), biological causality cannot be claimed.

Methods - Diets: The authors state the diets were "identical in cholesterol." A defining characteristic of many human Western diets is high cholesterol. While the matching controls for this variable, it limits the external validity of the "Western" label. The authors should explicitly acknowledge that this is a model of fat/protein source and fatty acid composition differences, rather than a "cafeteria-style" Western diet.

Methods - Antibiotics: There is no mention of exclusion criteria regarding antibiotic use during the 31-month trial. In long-term NHP studies, animals often require veterinary care. Were any subjects treated with antibiotics? If so, how was this handled in the microbiome analysis? This is critical for microbiome validity.

Methods - Statistical Analysis: The authors mention Pearson correlations. There is no mention of correction for multiple comparisons, despite correlating multiple metabolites, taxa, and phenotypes. Given the number of variables, false positives are a high risk.

Methods - Mediation: Please provide the specific software package or algorithm used for the "causal mediation analyses." Was bootstrapping used to estimate confidence intervals for the indirect effects? The sample size (n~19 per group) is extremely low for reliable mediation analysis.

Results: The removal of the Western diet subject based on "insulin physiology" and low pretreatment glucose (<100 mg/dL) seems arbitrary. <100 mg/dL is not necessarily pathological in macaques. Was this animal clinically hypoglycemic? Post-hoc removal of data points that disrupt correlations requires strictly defined, a priori exclusion criteria, otherwise it appears as p-hacking.

Results-Correlations: The correlations appear to be driven largely by the clustering of the two diet groups. For example, in Fig 4A, if you look only at the red dots (Western) or only at the blue dots (Mediterranean), is there still a significant slope? If the correlation exists only because Diet causes X and Diet causes Y, then X and Y are not necessarily mechanistically linked. The authors must report within-group correlations to validate the biomarker relationship.

Results - Mediation: The negative ACME for white matter (β=-0.309) is confusingly reported. Does this mean Oscillospira mediates the loss or the preservation? The directionality needs to be explicitly clear in the text.

Discussion - Butyrate: The text claims SCFAs like butyrate promote healthy aging, yet the Results (Line 349) explicitly state that butyrate did not differ between groups. The Discussion glosses over this. If Oscillospira is a butyrate producer (often cited as such), why was there no difference? This discrepancy needs to be reconciled.

Discussion - Mechanism: The authors speculate that SCFAs explain the benefit, but Acetate was the main driver, not Butyrate. The literature linking Acetate specifically to myelination is less robust than for Butyrate. The authors should expand on the specific role of Acetate in oligodendrocyte function if that is their proposed mechanism.

Discussion - Missing Function: The lack of cognitive testing is a major limitation. Preserved White Matter volume does not automatically equate to preserved function. The authors should be more transparent that this is a structural finding only.

Figure 5: The model implies Oscillospira -> Insulin Resistance -> White Matter. However, the text (Line 417) says the correlation between Insulin AUC and White Matter "did not reach statistical significance (p=0.058)." While marginal, building a causal model on a non-significant pathway is risky. The arrows in the diagram should perhaps be dashed or weighted to reflect statistical strength.

Data Availability: "Available upon request" is becoming less acceptable. The 16S sequencing data should be deposited in a public repository (NCBI SRA/ENA) with an accession number provided in the manuscript before publication.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

The authors have addressed the recommendations.

Author Response

Comments1: The authors have addressed the recommendations.

Response1: Thank you for your thoughtful review!

Reviewer 3 Report

Comments and Suggestions for Authors

1. The mediation framework requires temporal precedence (mediator measured between exposure and outcome), sufficient statistical power, and absence of unmeasured confounders between mediator and outcome. All three conditions are violated here. Oscillospira was measured at the end of the study (31 months), simultaneously with most outcomes, including white matter volume and sCD14. As such, the authors need to refrain from making claims that Oscillospira mediates anything when the "mediator" and "outcome" are measured at the same time point. The data are at best exploratory and associative.
2. The study uses 16S V4 sequencing processed through QIIME with OTU clustering at 97% identity and Greengenes-based taxonomy. That is an older pipeline with limited taxonomic resolution and known limitations. At this level of resolution, especially using OTUs and Greengenes, genus-level assignments can be uncertain, and the functional interpretation is even more speculative.
3. sCD14 is a nonspecific immune activation marker and can reflect several inflammatory processes; it is not a direct gut-barrier assay. Suggest to provide supporting references for its choice as well.
4. The study is presented as informing Alzheimer's disease risk but no AD biomarkers (amyloid, tau, neuroinflammation markers) were measured in this paper. Suggest to omit this please.

Comments on the Quality of English Language

Moderate edits required.

Author Response

Comment 1: The mediation framework requires temporal precedence (mediator measured between exposure and outcome), sufficient statistical power, and absence of unmeasured confounders between mediator and outcome. All three conditions are violated here. Oscillospira was measured at the end of the study (31 months), simultaneously with most outcomes, including white matter volume and sCD14. As such, the authors need to refrain from making claims that Oscillospira mediates anything when the "mediator" and "outcome" are measured at the same time point. The data are at best exploratory and associative.

Response 1: Thank you for this important point. In response to this concern, we have revised the manuscript to soften our language regarding mediation and removed statements implying causation. The analyses are also now described as exploratory and correlational, and we emphasize that these findings are hypothesis-generating rather than evidence of a mechanistic pathway. These clarifications have been incorporated throughout the Methods (lines 360-361), Results (lines 426, 443, 447) and Discussion (lines 533–572). We specifically note that future studies incorporating longitudinal sampling and temporally separated mediator and outcome measures (line 565) should be conducted to strengthen the work.

Additionally, we note that we have worked to thoughtfully incorporate and balance feedback from all four reviewers. While this process has substantially strengthened the manuscript, some reviewer suggestions were not entirely aligned. We have done our best to address these comments in a way that improves clarity and rigor while maintaining consistency across the manuscript, and we hope that the revisions satisfactorily address this concern.

 

Comment 2: The study uses 16S V4 sequencing processed through QIIME with OTU clustering at 97% identity and Greengenes-based taxonomy. That is an older pipeline with limited taxonomic resolution and known limitations. At this level of resolution, especially using OTUs and Greengenes, genus-level assignments can be uncertain, and the functional interpretation is even more speculative.

Response 2: Thank you for noting this limitation of the study. We have added language to our limitations section in the Discussion (lines 568-572), which underscores this point. The text reads, “Advances in technology (e.g., ASV-based pipelines and updated reference databases) also will help refine the taxonomic resolution, increasing certainty of genus-level assignments as well as making identification of species-level targets possible.”

 

Comment 3: sCD14 is a nonspecific immune activation marker and can reflect several inflammatory processes; it is not a direct gut-barrier assay. Suggest to provide supporting references for its choice as well.

Response 3: Thank you for noting this important point. We have revised the manuscript to better reflect the non-specific nature of sCD14 as a marker of immune activation (line 146). Additionally, we have clarified throughout the text that sCD14 reflects peripheral immune activation and is not a direct measure of gut barrier integrity (lines 291–293, 342–343, and 492–496). We also integrate supporting citations to justify its use as a biomarker associated with inflammatory activation (below):

Tabung, F.K.; Birmann, B.M.; Epstein, M.M.; Martínez-Maza, O.; Breen, E.C.; Wu, K.; Giovannucci, E.L. Influence of Dietary Patterns on Plasma Soluble CD14, a Surrogate Marker of Gut Barrier Dysfunction. Curr Dev Nutr 2017, 1, doi:10.3945/cdn.117.001396.

 

Cao, V.T.; Carter, M.C.; Brenchley, J.M.; Bolan, H.; Scott, L.M.; Bai, Y.; Metcalfe, D.D.; Komarow, H.D. sCD14 and Intestinal Fatty Acid Binding Protein Are Elevated in the Serum of Patients With Idiopathic Anaphylaxis. J Allergy Clin Immunol Pract 2023, 11, 2080-2086.e2085, doi:10.1016/j.jaip.2023.03.037.

 

Comment 4: The study is presented as informing Alzheimer's disease risk but no AD biomarkers (amyloid, tau, neuroinflammation markers) were measured in this paper. Suggest to omit this please.

 

Response 4: Excellent point. We have adjusted our language in the manuscript Introduction to better reflect our goals to explore the impacts of diet patterns on healthy versus pathological brain aging more generally, not specific to Alzheimer’s disease neuropathology (line 55-56). We also have added this citation:

 

Fekete, M.; Szarvas, Z.; Fazekas-Pongor, V.; Feher, A.; Csipo, T.; Forrai, J.; Dosa, N.; Peterfi, A.; Lehoczki, A.; Tarantini, S. Nutrition strategies promoting healthy aging: from improvement of cardiovascular and brain health to prevention of age-associated diseases. Nutrients 2022, 15, 47.

Reviewer 4 Report

Comments and Suggestions for Authors

Can be accepted.

Author Response

Comments 1: Can be accepted.

Response 2: Thank you for your thoughtful review!