title	author	date	output
Mouse Lingon diet experiment	name email affiliation Dmytro Kryvokhyzha & Shuyi Li [email protected] LUDC Bioinformatics Unit	19 maj, 2022	html_document keep_md toc true true

Publication

Kotowska et al. 2022 Short-term lingonberry feeding is associated with decreased insulin levels and altered adipose tissue function in high-fat diet fed C57BL/6J mice, Journal of Functional Foods

PI

Name: Karin Stenkula & Karin Berger

Email: [email protected]

Project

Mice were fed with different diet for 4 days:

• HFD - high fat diet,
• LFD - low fat diet,
• Lingon - high fat diet with Lingon berries.

HFD vs Lingon is the most interesting comparison. LFD is a control.

RNA-seq data were collected from adipose tissue.

Each group has 5 samples.

Analysis steps:

1. QC sequencing data
2. Map and count reads
3. Make pair-wise comparisons between three groups (HFD vs LFD/HFD vs Lingon/LFD vs Lingon) to find the differential expressed (DE) genes
4. Perform enrichment analysis (GO & KEGG)
   • Gene set enrichment analysis
   • Over-representation analysis
5. Check interesting gene, GO terms and KEGG pathways
   • Potential interesting genes
     • FABP4
     • Fsp27
     • CIDEA
     • SLC2A4
     • PNPLA2
     • PLIN1
     • CAV1
     • PPARG
     • DGAT2
     • Anxa2
     • Aacs
     • Acacb
     • Acly
     • Elovls
     • Acots
     • Cidec
     • Insig2
     • ApoA4
   • Interesting GO
     • glucose metabolism
     • adipogenesis
     • mitochondria functions
   • Interesting pathways
     • Lipid synthesis
     • Regulation of lipid metabolism
     • Enzymes in fatty acid activation and oxidation
     • Lipolysis
     • Cholesterol metabolism
     • Glucose uptake

Data

The RNA-Seq data is uploaded to the NCBI under the BioProject PRJNA783912.

All the data is located on the Indigo server: /ludc/Active_Projects/Mouse_Adipocite_Lingonberry/ludc/

Prerequisites

You need to install Conda and load the pre-configured conda environment. It should also install all the required programs.

conda env create -f conf/conda.yml
conda activate LingonProj

Analysis

QC

Preliminary fastq QC results can be found in: ~/results/tables/multiqc/

QC reports can be found in: ~/results/reports/

Map and count reads

Performed with Salmon:

snakemake -s scr/salmon.smk \
   -j 100 \
   -p --use-conda \
   --cluster-config conf/cluster.yml \
   --cluster "condor_qsub -o logs/{rule}.out -e logs/{rule}.err -l procs={cluster.cores},mem={cluster.ram} -m e -V"

Results:

• results/tables/salmon/{sample}/quant.sf.gz - transcripts counts
• results/tables/salmon/{sample}/quant.genes.sf.gz -gene-level counts

Differential expression

Performed with DESeq2:

R -e 'rmarkdown::render("scr/DESeq.Rmd", output_dir="results/reports/")'

Results:

• results/reports/DESeq.html - notebook describing the analysis
• results/tables/deseq/LingonProj_DESeqres.csv - differential expression results with TMP (three results in one table).
• results/figures/ - plots saved as pdf files

Enrichment analysis

Performed with WEB-based GEne SeT AnaLysis Toolkit:

R -e 'rmarkdown::render("scr/WebGestaltR.Rmd", output_dir="results/reports/")'

Results:

• results/reports/Project_HFD_vs_Lingon_FDR_0_1_ORA - ORA analysis of HFD_vs_Lingon FDR < 0.1.
• results/reports/Project_HFD_vs_LFD_FDR_0_01_ORA - ORA analysis of HFD_vs_LFD FDR < 0.01.
• results/reports/Project_LFD_vs_Lingon_FDR_0_01_ORA - ORA analysis of LFD_vs_Lingon FDR < 0.01.

Select interesting gene

Note：interesting_gene.txt is created in advance (case insensitive)

python scr/select_gene.py -i results/tables/deseq/LingonProj_DESeqres.csv -g data/reference/interesting_gene.txt -o results/tables/deseq/LingonProj_interesting_gene.csv

Results:

• results/tables/deseq/LingonProj_interesting_gene.csv - DESeq results with only interesting gene (listed above in the analysis step) selected.

Compare with the published results

We compare our results with the paper Intact glucose uptake despite deteriorating signaling in adipocytes with high-fat feeding.

We expect the highest correlation of our results with 4 days results in this paper.

R -e 'rmarkdown::render("scr/compare_with_published.Rmd", output_dir="results/reports/")'

We see the similarity in fold change and day 4 factor in ANOVA, but mean expression is also similar to day 2 and sometimes to day 6 depending at what genes we look. This may be because the food is a little different - the high fat diet with Lingon contained 45% fat, compared with my previous short-term HFD study where the diet contained 58% fat.

Lab verification

We decided to verify in the lab the following genes:

• mitochondrial fission
• angiogenesis

These genes are selected in the scr/WebGestaltR.Rmd notebook and output to:

• results/tables/candidates_lab_verification/HFD_vs_Lingon_mitochondrial_fission.csv - DESeq results for genes related to mitochondrial fission.
• results/tables/candidates_lab_verification/HFD_vs_Lingon_mitochondrial_fission.pdf - expression of genes related to mitochondrial fission.
• results/tables/candidates_lab_verification/HFD_vs_Lingon_angiogenesis.csv - DESeq results for genes related to angiogenesis.
• results/tables/candidates_lab_verification/HFD_vs_Lingon_angiogenesis.pdf - expression of genes related to angiogenesis.

Maqui and Lingon overlap

One of the reviewers requested to check the overlap between gene affected by maqui berry and our results.

I manually extracted the list of genes from Fig.3 in https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6769892/ and extracted these genes from our results:

head -n 1 results/tables/deseq/LingonProj_DESeqres.csv \
   > results/tables/Maqui_lingon_genes_overlap.csv
grep -f data/Maqui_genes.txt results/tables/deseq/LingonProj_DESeqres.csv \
   >> results/tables/Maqui_lingon_genes_overlap.csv

Only Acaca, Ppargc1a, Acly, Fasn were among the significant results in the comparison HFD_vs_LFD. These genes are up-regulated in LFD. Acaca, Ppargc are significant in the comparison LFD_vs_Lingon and up-regulated in LFD.

There are no significant matches in HFD_vs_Lingon, although Acaca has a p-value of 0.136 and it is up-regulated in Lingon.

So, we may say that we see similar effect only for the Acaca gene that is up-regulated by both Maqui and Lingon berries. We can also mention that although we do not get significant difference for these genes in the comparison HFD_vs_Lingon, all these genes except Prdm16, Cpt1b, Acox3, Prdm16 are expressed on average at higher level in Lingon than in HFD. This overlap is significantly non-random:

dat <- data.frame(
  "Maqui" = c(15, 0),
  "Lingon" = c(11, 4),
  row.names = c("up", "down"),
  stringsAsFactors = FALSE
)
dat
fisher.test(dat, alternative = "greater")

So, the effect seems to be partially similar between Maqui and Lingon berries.