Search Weight Loss Topics:

Home » Uncategorized » Diet-omics in the Study of Urban and Rural Crohn disease Evolution (SOURCE) cohort – Nature.com

Study population

The study was approved by ethics committees in both Sheba Hospital in Israel (No. 5484) and the First-affiliated hospital of Sun Yat-Sen (SYS) University in China (No.[2021]GH1457; No.[2019]GH0367; No.[2017]073), and the research complies with all relevant ethical regulations. This study was conducted in Guangdong province in Southern China and Israel (Jan 2019 to April 2021). Several populations were analyzed (Fig.1). In China, newly diagnosed and treatment-nave CD patients were included, along with healthy urban residents of Guangzhou, a modernized metropolitan community with a population of 16 million, and healthy residents of Shaoguan district, a rural underdeveloped community 300km north of Guangzhou. Participants were asked about the amount of time they spent in an urban environment in the last year (How long have you stayed in a city in the last year?) with answers including less than 10%, 1050%, and above 50%. Newly diagnosed CD patients and healthy controls from Israel were included as another layer of a Westernized control cohort. CD diagnoses of all patients were harmonized and based on clinical history, physical examination, laboratory work, radiological findings, and endoscopic and histological features as previously established in the European Crohns and Colitis Organization consensus statement36. Written informed consent was obtained from all participants.

Data regarding enrolled subjects were recorded in a structured manner that included demographic, clinical, laboratory, endoscopic, and pathological features for the indicated group and participants (Fig.1). Laboratory tests included C-reactive protein (CRP). Endoscopic evaluations included gut segmental involvement. Stool specimens were collected into a collection tube at least 3 weeks following any antibiotic treatment. Stool samples were aliquoted and frozen immediately in 80C. Ileal biopsies were gathered during diagnostic colonoscopy and stored in RNAlater and frozen at 80C. Due to COVID-19 outbreak limitations, samples were processed locally within each country of origin. Samples from both cohorts were analyzed using a similar bioinformatic pipeline; direct comparisons between groups were performed within each country of origin.

Patients from both cohorts underwent environmental and dietary exposure surveys. Gender was determined based on self-report. For environmental exposure, we used the questionnaire developed by the International Organization of IBD (IOIBD), with some modifications. The questionnaire consists of 87 questions covering 25 different topics proposed to be environmental risk factors for CD. Although it was not formally validated by the IOIBD, this questionnaire has been previously used in epidemiological studies investigating triggers of IBD37,38, including one conducted in South-East Asia and China39. IOIBD questions relate to five main different areas: (1) Childhood factors up to 20 years including breastfeeding, appendectomy, tonsillectomy, eczema, vaccinations (tuberculosis, pertussis, measles, rubella, diphtheria, tetanus, polio), childhood infections (measles, pertussis, rubella, chickenpox, mumps, scarlet fever) and pet ownership; (2) food habits including daily, weekly or less frequent consumption of fruit, vegetables, egg, cereal, bread, coffee, tea, juice, sugar, and fast food; (3) smoking habits (current smoker, non-smoker, ex-smoker); (4) sanitary conditions such as the availability of in-house water tap, hot water tap or flush toilet; and (5) others factors including daily physical activity, the oral contraceptive pill and stressful events before diagnosis. We also included in the IOIBD questionnaire items about antibiotic use before and after the age of 15 years, use of toothpaste, and the presence of amalgam teeth filling during childhood or later in life. In addition, we added a specific question specifying the amount of time spent in an urban environment in the last year with answers including less than 10%, 1050%, and above 50%. Although the IOIBD questionnaire explores the role of diet before the diagnosis of IBD, we added an additional comprehensive FFQ (Food frequency questionnaire), conducted by a trained dietician. This tool included over 600 food items, with an FFQ list that prioritized the foods/beverages accounting for at least 80% of the total energy intake and the between-person variance in previously collected dietary intake data from the adult Israeli population40,41. Personnel from both Sheba and SYS were similarly trained in the dietary interview method and the equivalent FFQ was used after translation to Chinese and adaptation to the Chinese diet. In Israel, a computerized FFQ version was used that automatically computes the average daily nutrient content of an individual patients diet including macro and micronutrients, and food item servings. In China, the data were extracted manually and the portion of a specific food, and micro and micronutrient consumption were summarized. Data regarding exposures and diet are summarized in Supplementary Dataset1. The Sheba FFQ output included 62 features, while the SYS one included 28 features.

Principal component analysis (PCA) was performed to summarize variation in Israel and China FFQ data, separately, using R prcomp function with data scaled and centered. FFQ components were correlated to PC1 and PC2 values using Spearmans correlation, and the top 10 FFQ features by maximal Spearmans rho correlation to PC1 or PC2, with p0.05, are shown to highlight the main components affecting variance in dietary intake. Spearmans correlation was calculated between all FFQ component pairs, separately for Israel, China rural, and China urban and CD. Correlation heatmaps were generated showing Spearmans rho values, clustered using R hclust function with Euclidian distances.

At the Sheba site (Israel), fecal42,43,44 and biopsy7,8,9,10 DNA extraction, and PCR amplification of the variable region 4 (V4) of the 16S rRNA gene using Illumina adapted universal primers 515F/806R was conducted using the direct PCR protocol [Extract-N-Amp Plant PCR kit (Sigma-Aldrich, Inc.)]42,43,44. At the SYS site (China), fecal DNA samples were extracted using the OMEGA Soil DNA Kit (Omega Bio-Tek, Norcross, GA, USA) following the manufacturers instructions. PCRs of the variable region 4 (V4) of the 16S rRNA were conducted and amplicons were pooled in equimolar concentrations into a composite sample that was size selected (300500bp) using agarose gel to reduce non-specific products from host DNA. Sequencing was performed on the Illumina MiSeq platform at Sheba or the NovaSeq platform at Shanghai Personal Biotechnology Co., Ltd (Shanghai, China). Samples from both cohorts were analyzed independently using a similar bioinformatic pipeline. Reads were processed in a data curation pipeline implemented in QIIME 2 version 2021.445,46. Reads were demultiplexed according to sample-specific barcodes. Quality control was performed by truncating reads after three consecutive Phred scores lower than 20. Reads with ambiguous base calls or shorter than 150bp after quality truncation were discarded. Amplicon sequence variants (ASVs) detection was performed using Deblur47, and duplicate samples from different runs were joined, resulting in 323 samples with a median of 55,844 reads/sample (IQR 50,43562,947) for China, and 158 samples with a median of 23,595 reads/sample (IQR 13,54936,505) for Israel. ASVs present in less than 1% of the samples were removed. Additionally, candidate contaminant ASVs were filtered using dbBact14 by removing ASVs with the f-score mean for (water, soil, mus musculus) higher than homo sapiens, resulting in 1668/3642 ASVs for China and 1290/2838 ASVs for Israel after filtering. ASV taxonomic classification was assigned using a naive Bayes fitted classifier, trained on the August 2013 Greengenes database. Taxonomy assigned by 16S is indicated by the specific ASV number, and the sequence associated with each ASV number is indicated in Supplementary Dataset2 and in the relevant supplementary datasets. All samples were rarefied to 33k reads for and diversity analysis for China samples, and 4k reads for Israel samples, to avoid read number effects. Faiths phylogenetic diversity48 was used as a measure of within sample diversity, and Unweighted UniFrac was used as a measure of between sample -diversity49, using a phylogenetic tree generated by SAT-enabled phylogenetic placement (SEPP)50. The resulting distance matrix was used to perform a Principal Coordinates Analysis (PCoA). Heatmaps were generated using Calour version 2018.10.1 with default parameters51.

PERMANOVA: quantifications of variance were calculated using PERMANOVA with the adonis function in the R package Vegan52, on the rarefied Unweighted UniFrac distance values. The total variance explained by each variable was calculated while accounting for age and gender in the model (except for when looking at the contribution of age and gender, when only age or gender can be controlled for). PERMANOVA was calculated independently for each group (China CD, urban, rural-urban and rural, and Israel CD and control) and for each questionnaire and FFQ component. Multivariate Association with Linear Models (MaAsLin2) was used with R package version 1.8.0, to test for specific differentially abundant ASVs between: rural and rural-urban samples controlling for age and gender, China CD and urban controls controlling for age and gender, and Israel CD and controls, using both stool and biopsy samples, controlling for age, gender, sample type (stool or biopsy) and patient ID as the random variable. MaAsLin2 was also used to identify ASVs correlated with dietary consumption of fat, iron, and protein separately, within all China controls, controlling for age, gender and group (urban, rural-urban or urban). A false discovery rate (FDR) cutoff of 0.25 was used for all MaAsLin2 analysis53, and FDR cutoff of 0.1 is indicated.

Rural and Health indices: per-sample health index was calculated as previously described11. Briefly, a set of ASVs that were significantly increased or decreased across multiple human diseases compared to controls was identified. Using these ASVs bacteria, for each sample the log of the ratio of health-associated bacteria (98) to disease-associated bacteria (32), following rank transforming the samples, was calculated and defined as the health index (with higher values indicating a better health-associated microbiome). A similar approach was used to define a rural index for each sample, as follows: Using an independent dataset of rural and urban Chinese samples5, we identified 76 and 42 ASVs significantly higher/lower in the rural community respectively (using a rank-mean test with dsFDR<0.154, implemented in Calour51). The rural index was then calculated for each SOURCE sample as the log of the ratio of the rank-transformed frequencies of the ASVs from the rural and urban ASVs. Age matching between sample groups was performed by binning ages into 10 year bins, and equalizing the number of samples in each age bin between the two sample groups by randomly dropping samples.

dbBact term enrichment analysis: significantly enriched dbBact14 ontology terms between two ASV sets (e.g., higher/lower in rural community or positively/negatively correlated with dietary factor) were identified using the dbBact-calour plugin. Briefly, dbBact contains annotations linking ASVs to ontology-based terms, based on manual analysis of over 1000 amplicon experiments. For the current experiment analysis, for each term, a dbBact annotation-based score is calculated for each ASV, and the distribution of the score across the two ASV groups is compared to random permutations (of ASV group labels), using a permutation-based rank-mean test with dsFDR multiple hypothesis correction. For the term-specific Venn diagrams, the number of ASVs associated with the term in at least one dbBact experiment is shown for each ASV group, with the central (gray) circle showing the total number of ASVs in dbBact associated with the term. The study was conducted and is currently reported according to the STORMS guidelines55 (information in Supplementary Dataset1).

For samples from the Sheba site, DNA was purified using the PowerMag Soil DNA isolation kit (MoBio) optimized for the Tecan automated platform. DNA was diluted to 1.5ng, and Illumina libraries were prepared using Nextera DNA library preparation kit, Ref# 15028211; by Tecan Freedom Evo 200 robot device. Nextera DNA Unique Dual Indexes Sets AD from IDT were used for library preparation. Library concentration was measured using the iQuantTM dsDNA HS Assay Kit, ABP biosciences (Cat# AP-N011), and library size was quantified by automated electrophoresis nucleic acid QC -Tape-Station system. Libraries were sequenced by a NextSeq 500 device with IlluminaNS 500/550 High Output V2 75 cycle kit, Cat# FC-404-2005. SYS samples site, the extracted DNA was processed to construct metagenome shotgun sequencing libraries with insert sizes of 400bp by using Illumina TruSeq Nano DNA LT Library Preparation Kit. Each library was sequenced by the Illumina HiSeq X-ten platform (Illumina, USA) with PE150 strategy at Personal Biotechnology Co., Ltd. (Shanghai, China). Samples from both cohorts were analyzed independently using a similar pipeline (https://github.com/biobakery). Reads were first decontaminated and trimmed using KneadData v0.12.0, then samples under 7M reads PE (or 3.5M SE) were excluded. The average number of reads after the quality control process was 52,508,783.72 (11,833,357.89) PE and 8,773,250.37 (2,135,013.94) SE for the Chinese and Israeli cohorts respectively. Taxonomic profiles were generated using MetPhaln v4.0.056, from which the functional profiles were generated using HUMAnN v3.657. Default parameters were used for all modules, besides defining concatanating PE reads using the -cat-final-output parameter in KneadData. Taxonomic and functional features were filtered out if they didnt have abundance greater than 0.01% in at least 10% of the samples.

For the Sheba samples (n=37), extraction solution (ES: 75% methanol and 25% water and six internal standards) was mixed with fecal smear, sonicated for 10min, centrifuged at 14,000g for 10min at 4C, and stored at 80C until submission for LC-MS metabolomics analysis. LC-MS analysis was conducted as described58. Briefly, Dionex Ultimate ultra-high-performance liquid chromatography (UPLC) system coupled to Orbitrap Q-Exactive Mass Spectrometer (Thermo Fisher Scientific) was used. The resolution was set to 35,000 at a 200 mass/charge ratio (m/z) with electrospray ionization and polarity switching mode to enable both positive and negative ions across a mass range of 671000m/z. The UPLC setup consisted of ZIC-pHILIC column (SeQuant; 150mm2.1mm, 5m; Merck). Stool extracts were injected, and the compounds were separated with mobile phase gradient, starting at 20% aqueous (20mM ammonium carbonate adjusted to pH 9.2 with 0.1% of 25% ammonium hydroxide) and 80% organic (acetonitrile) and terminated with 20% acetonitrile. Flow rate and column temperature were maintained at 0.2ml/min and 45C, respectively, for a total run time of 27min. All metabolites were detected using mass accuracy below 5ppm. Thermo Xcalibur 4.1 was used for data acquisition. Peak areas of metabolites were determined using MZmine2.5359 by using the exact mass of the singly charged ions (m/z) and the retention time of metabolites was predetermined on the pHILIC column by analyzing an in-house mass spectrometry metabolite library that was built by running commercially available standards. Thermo TraceFinderTM 4.1 software was used for validation, by comparing the peak areas of the internal standards determined by both software. A total of 405 of 545 of the predefined metabolites library passed the threshold of peak intensity and were included in the analyses.

For the SYS samples, a targeted metabolomic analysis using a Q300 kit (Metabo-Profile, Shanghai, China) was performed. Lyophilized samples (~5mg) were mixed with 25l of water and homogenized with zirconium oxide beads for 3min. One hundred twenty l of methanol containing internal standard was added and then homogenized for another 3min, centrifuged at 18,000g for 20min, and 20l of supernatant was transferred to a 96-well plate. The plate was sealed and incubated at 30C for 60min, after which 330l of ice-cold 50% methanol solution was added to dilute the sample. Then the plate was stored at 20C for 20min, followed by 4000g centrifugation at 4C for 30min. One hundred thirty-five l of supernatant was transferred to a new 96-well plate with 10l internal standards in each well. A liquid chromatography coupled to tandem mass spectrometry (UPLC-MS/MS) system (ACQUITY UPLC-Xevo TQ-S, Waters Corp., Milford, MA, USA) was used to quantitate all targeted metabolites. To diminish analytical bias within the entire analytical process, the samples were analyzed in duplicates that were randomly analyzed. The quality control (QC) samples, internal standard calibrators, and blank samples were analyzed across the entire sample set. The raw data files generated by UPLC-MS/MS were processed using the MassLynx software (v4.1, Waters, Milford, MA, USA) to perform peak integration, calibration, and quantitation for each metabolite. A total of 185 of 305 metabolites passed the threshold of peak intensity and were included in the analyses.

Ninety-two metabolites overlapped between the 185 SYS dataset and the 405 in the Sheba cohort, and these were used to test correlations as independent validation. Overall, the normalized metabolite levels (each metabolite value was divided by the sum of total metabolites value per sample) were used for all downstream analyses in both Sheba and SYS cohorts.

Principal coordinates analysis (PCoA) was performed on the metabolomics data using Canberra distances as a measure of between sample -diversity. Multivariate Association with Linear Models (MaAsLin2) was used with R package version 1.8.0, to test for specific differentially abundant metabolites between China rural and rural-urban samples controlling for age and gender, and for China CD and urban controls controlling for age and gender, using FDR cutoff of 0.25.

At the Sheba site, RNA and DNA were isolated from terminal ileum (TI) biopsies obtained during diagnostic colonoscopy using the Qiagen AllPrep RNA/DNA Mini Kit. PolyA-RNA selection, fragmentation, cDNA synthesis, adapter ligation, TruSeq RNA sample library preparation (Illumina, San Diego, CA), and paired-end 75bp sequencing were performed. Median reads depth was ~ 39M (3146M IQR) in Sheba. Samples were sequenced at the NIH -supported Cincinnati Childrens Hospital Research Foundation Digestive Health Center. At the SYS site, mRNA was purified from total RNA using poly-T oligo-attached magnetic beads. The strand-specific cDNA sequencing libraries were generated using NEBNext UltraTM Directional RNA Library Prep Kit for Illumina (NEB, USA), and index codes were added. Samples were purified (AMPure XP system) and clustering of the index-coded samples was performed on a cBot Cluster Generation System using HiSeq 4000 PE Cluster Kit (Illumina, NEB, USA). After cluster generation, the stranded, poly-A selected libraries were sequenced on an Illumina NovaSeq 6000 platform by Novogene Bioinformation Technology. One hundred fifty bp paired-end reads were generated to a median depth of 42.3M (39.946.6M IQR) reads for China samples, and 38.2M (3238.2M IQR) reads for Israel samples. Reads were quantified by kallisto60 version 42.5 using Gencode v24 as the reference genome. Kallisto output files were summarized to gene level using R package tximport version 1.22.061. Protein coding genes with Transcripts per Million (TPM) values above 1 in at least 20% of the samples were used in downstream analysis.

Weighted gene co-expression network analysis (WGCNA) to identify modules of co-expressed genes25,62 was implemented utilizing R WGCNA package version 1.72-1, using the Sheba cohort as described63,64. The analysis was performed on TI transcriptomics data. WGCNA was implemented for the identification of co-expressed gene clusters; we used pairwise correlations between gene expression profiles and the signed hybrid version of WGCNA. Similarities of gene co-expression are converted to adjacency values (power adjacency function), with a parameter of 12. Average linkage hierarchical clustering on TOM-based dissimilarities is implemented to detect modules of strongly correlated genes across samples. For each module, the first principal component, referred to as the eigengene, was considered to be the module representative. A module summarized the expression levels of all the genes in a given module. Included parameters were cluster sensitivity parameter (deepSplit) of 2 to identify balanced gene modules, a minimum number of genes in a module (minModuleSize) was set to 30 genes, maxBlockSize was set to 20,000 to include all genes in one block. The same modules of co-expressed genes were applied to the SYS transcriptomics dataset. We focused on modules significantly associated with disease, with p0.05 in at least one of the two cohorts. Nine out of 14 modules were associated with disease. Module eigengenes were additionally correlated (Students asymptotic p value) to dietary factors, and to fecal metabolites. Metabolomics data was log-scaled and cleaned, with zeroes replaced by a fifth of the lowest value per metabolite, and values with over 4 standard deviations from the mean were trimmed to 4 standard deviations from the mean, to avoid extreme values driving correlations. BenjaminiHochberg FDR correction was applied separately to diet and metabolomics results. Correlations with FDR0.25 were considered significant. This module eigengenes correlation analysis was performed independently for the Sheba and SYS datasets. ToppGene65 and ToppCluster software were used to perform Gene Set Enrichment Analyses (GSEA) of the protein-coding genes within the modules in the WGCNA TI analysis.

For each of the Sheba TI transcriptomics WGCNA disease-associated modules, we defined modules associated with metabolites as fecal metabolites with significant correlation (FDR0.25) to that module eigengene. An exact binomial test was used to test the consistency of metabolites correlation to TI WGCNA modules, between all samples and CD samples only in the Sheba cohort. For the 416 significant module-metabolites correlation calculated using all samples, a correlation was considered consistent if it changed in the same direction in CD samples, looking at Spearmans rho. 406 out of 416 correlations were consistent. The probability of success was calculated as pall*pcd+(1pall)*(1pcd), with pall representing the percentage of positive correlations in all samples, and pcd representing the percentage of positive correlations in CD samples, to account for the unbalanced positive to negative correlations ratio. HAllA (Hierarchical All-against-All significance testing) version 0.8.20 was used to identify potential correlations between these modules associated metabolites and Israeli FFQ components, stool 16S ASVs, and stool metagenomics (MGX species, pathways, and ECs). HAllA was used with Spearman correlation and FDR cutoff of 0.25.

We applied sparse Partial Least Squares (sPLS) regression between pairs of omics19 including -metagenomics (species, pathways, and ECs), stool metabolomic, and host TI transcriptomics (using disease-related WGCNA modules PC1 values). This method aims to maximize the shared variation between a pair of omics while accounting for the sparsity in the data. As a measure of the identified shared variation, we calculated the Spearman correlation between the first components of each omic. Significance was evaluated by generating 100 permutations of the feature table, and counting the number of permutations that yielded a higher correlation value than the one calculated for the original data. DIABLO20 (Data Integration Analysis for Biomarker discovery using Latent variable approaches for Omics studies) was used to simultaneously maximize the shared variation between each omics and differentiate between the health conditions. To uncover the relationship between the omics features and the disease state, we focused on the 10 loadings with the highest correlation to the first and second components. Both sPLS regression and DIABLO were calculated using the MixOmics package66. According to the recommendations in the MixOmics package, the number of components was chosen by a minimal balanced error rate using the centroid distance method. In addition, HAllA version 0.8.20 was used to identify potential correlations between China rural and rural-urban metabolites, FFQ components, and stool 16S ASVs, using Spearman correlation and FDR cutoff of 0.25.

Statistics used for transcriptomics, microbiome, and metabolomics were performed in R, and details are under these specific sections. Overall, Pearsons chi-square test or Fishers exact test was used for categorical variables, Spearmans rank correlation was used for continuous variables, and the MannWhitney U test for categorical variables, with BenjaminiHochberg Procedure for FDR correction using 0.1 or 0.25 as indicated.

Further information on research design is available in theNature Portfolio Reporting Summary linked to this article.

See more here:
Diet-omics in the Study of Urban and Rural Crohn disease Evolution (SOURCE) cohort - Nature.com

Your Full Name
Your Email
Your Phone Number
Select your age (30+ only)	30313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120
Select Your US State	Select Your StateAlabamaAlaskaArizonaArkansasCaliforniaColoradoConnecticutDelawareFloridaGeorgiaHawaiiIdahoIllinoisIndianaIowaKansasKentuckyLouisianaMaineMarylandMassachusettsMichiganMinnesotaMississippiMissouriMontanaNebraskaNevadaNew HampshireNew JerseyNew MexicoNew YorkNorth CarolinaNorth DakotaOhioOklahomaOregonPennsylvaniaRhode IslandSouth CarolinaSouth DakotaTennesseeTexasUtahVermontVirginiaWashingtonWashington,DCWest VirginiaWisconsinWyoming
Program Choice	Select Your ProgramInjectable HGHInjectable TestoteroneInjectable SermorelinInjectable Weight LossInjectable HCG DietInjectable Vitamins
Confirm over 30 years old	Yes
Confirm that you resident in USA	Yes
This is a Serious Inquiry	Yes
Message:

Home » Uncategorized » Diet-omics in the Study of Urban and Rural Crohn disease Evolution (SOURCE) cohort – Nature.com