2025-05-08 13:57:45 -04:00
|
|
|
# Research Aims
|
|
|
|
|
|
|
|
## 1. Ybx1 in adipogenesis
|
|
|
|
### 1.1 Identifying YBX1-CEBPA-cBAF interactions in early, middle, late 3T3-L1 adipogenesis
|
|
|
|
- **Hypothesis:** YBX1 cooperates with CEBPA and cBAF complex to regulate temporal gene expression during adipogenesis
|
|
|
|
- **Approach:** Time-course analysis of 3T3-L1 differentiation (days 0, 2, 4, 6, 8)
|
|
|
|
- ChIP-seq for YBX1, CEBPA, BRG1, SMARCD2, SMARCE1
|
|
|
|
- RNA-seq to correlate binding with expression changes
|
|
|
|
- Co-IP followed by MS to identify temporal protein-protein interactions
|
|
|
|
- **Expected outcomes:** Map dynamic transcriptional networks controlling adipocyte differentiation
|
|
|
|
- **Question:** Are we seeing this in all cell types?
|
|
|
|
|
|
|
|
### 1.2 YBX1 in adipogenic 3T3 metabolism
|
|
|
|
- **Hypothesis:** YBX1 regulates metabolic reprogramming during adipogenesis
|
|
|
|
- **Approach:**
|
|
|
|
- Metabolic profiling of control vs YBX1-depleted 3T3-L1 cells during differentiation
|
|
|
|
- Seahorse analysis of glycolysis and mitochondrial function
|
|
|
|
- Lipidomics to characterize changes in lipid composition
|
|
|
|
- Integration with transcriptomic data to identify YBX1-dependent metabolic pathways
|
|
|
|
- **Expected outcomes:** Define YBX1's role in adipocyte metabolic adaptation
|
|
|
|
|
|
|
|
### 1.3 Loss of SMARCD2, SMARCE1, and YBX1 in differentiating 3T3 cells
|
|
|
|
- **Hypothesis:** cBAF subunits SMARCD2, SMARCE1 cooperate with YBX1 to maintain proper adipogenic program
|
|
|
|
- **Approach:**
|
|
|
|
- CRISPR/shRNA knockdown of individual and combined factors
|
|
|
|
- Oil Red O staining quantification
|
|
|
|
- RNA-seq and ATAC-seq to identify chromatin and expression changes
|
|
|
|
- Rescue experiments with reconstituted factors
|
|
|
|
- **Expected outcomes:** Mechanistic understanding of how chromatin remodeling complex and YBX1 coordinate adipogenesis
|
|
|
|
|
|
|
|
## 2. cBAF-CEBPa regulating mir-101
|
|
|
|
- **Hypothesis:** cBAF and CEBPA coordinate to regulate miR-101 expression, impacting lipid metabolism
|
|
|
|
- **Approach:**
|
|
|
|
- ChIP-seq for cBAF components and CEBPA at miR-101 locus
|
|
|
|
- miR-101 overexpression and inhibition studies
|
|
|
|
- Target validation using directional RNA-seq and ribosome profiling
|
|
|
|
- Lipidomics to assess impact on lipid composition
|
|
|
|
- **Expected outcomes:** Novel miRNA-mediated mechanism in metabolic control
|
|
|
|
|
|
|
|
## 3. Ybx1 in metabolic reprogramming of hepatocytes
|
|
|
|
### 3.1 Transcriptional network analysis of PPARg, CEBPa, CEBPb, SMARCD2, SMARCE1, BRG1
|
|
|
|
- **Hypothesis:** Chronic fat exposure alters binding patterns of master regulators and chromatin remodelers
|
|
|
|
- **Approach:**
|
|
|
|
- Primary hepatocytes and HepG2 cells exposed to different lipid conditions (acute vs chronic)
|
|
|
|
- ChIP-seq for all factors
|
|
|
|
- ATAC-seq to assess chromatin accessibility changes
|
|
|
|
- Integration with RNA-seq to identify dysregulated pathways
|
|
|
|
- **Expected outcomes:** Map how fat exposure reshapes the regulatory landscape in hepatocytes
|
|
|
|
|
|
|
|
### 3.2 Posttranscriptional regulation of lipogenic genes by Ybx1
|
|
|
|
- **Hypothesis:** YBX1 regulates mRNA stability and translation of key lipogenic genes
|
|
|
|
- **Approach:**
|
|
|
|
- RIP-seq for YBX1-bound mRNAs
|
|
|
|
- mRNA decay assays in control vs YBX1-depleted cells
|
|
|
|
- Polysome profiling coupled with RNA-seq
|
|
|
|
- CLIP-seq to map direct YBX1-RNA interactions
|
|
|
|
- **Expected outcomes:** Novel posttranscriptional regulatory mechanism in lipid metabolism
|
|
|
|
|
|
|
|
### 3.3 YBX1 INS1 repression
|
|
|
|
- **Hypothesis:** YBX1 represses insulin signaling in non-pancreatic tissues
|
|
|
|
- **Approach:**
|
|
|
|
- Cell-type specific analysis of YBX1 and INS1 expression
|
|
|
|
- Reporter assays with INS1 promoter
|
|
|
|
- YBX1 ChIP-seq in pancreatic vs non-pancreatic cells
|
|
|
|
- CRISPR activation/repression to modulate YBX1 levels
|
|
|
|
- **Key questions:**
|
|
|
|
- Is YBX1 absent in pancreatic beta cells? If so, how?
|
|
|
|
- Can targeting YBX1 be therapeutic for certain types of Type 1 diabetes?
|
|
|
|
|
|
|
|
### 3.4 YBX1-CEBPa-cBAF in lipid-exposed hepatocytes multiomics
|
|
|
|
- **Hypothesis:** Chronic lipid exposure alters the cooperative activity of YBX1-CEBPa-cBAF, leading to hepatic steatosis
|
|
|
|
- **Approach:** Comprehensive multi-omic analysis integrating:
|
|
|
|
- mRNA-seq
|
|
|
|
- ATAC-seq
|
|
|
|
- ChIP-seq: YBX1, BRG1, CEBPa, CEBPb
|
|
|
|
- CelSeq2 for single-cell resolution of heterogeneous responses
|
|
|
|
- DUO LINK proximity assays to validate physical interactions
|
|
|
|
- Lipidomics to correlate with phenotypic outcomes
|
|
|
|
- **Expected outcomes:** Systems-level understanding of transcriptional dysregulation in fatty liver development
|
|
|
|
|
|
|
|
## 4. Using artificial intelligence to enhance biological insights
|
|
|
|
### 4.1 Geo stacking app
|
|
|
|
- **Project goal:** Develop tool to integrate and visualize multiple GEO datasets
|
|
|
|
- **Potential applications:** Meta-analysis of metabolic disease datasets
|
|
|
|
|
|
|
|
### 4.2 Transregulator-ATAC pattern finder
|
|
|
|
- **Project goal:** Machine learning tool to predict transcription factor binding from ATAC-seq data
|
|
|
|
- **Approach:** Train models using paired ATAC-seq and ChIP-seq datasets
|
|
|
|
|
|
|
|
### 4.3 Increase read mapping speed
|
|
|
|
- **Project goal:** Optimize computational pipeline for multi-omic data analysis
|
|
|
|
|
|
|
|
### 4.4 Lab agent
|
|
|
|
- **Project goal:** Develop AI-assisted laboratory workflow management system
|
|
|
|
- **Applications:** Experiment planning, protocol optimization, data analysis
|
|
|
|
|
|
|
|
### 4.5 ML model for oil-red O image quantification
|
|
|
|
- **Project goal:** Automated quantification of lipid accumulation in cell culture
|
|
|
|
- **Approach:** Computer vision models trained on labeled microscopy images
|
|
|
|
|
|
|
|
## 5. Regulation of Ybx1 cyto-nuclear translocation
|
|
|
|
- **Hypothesis:** Nutrient status regulates YBX1 localization and function
|
|
|
|
- **Approach:**
|
|
|
|
- Subcellular fractionation followed by western blot
|
|
|
|
- Live-cell imaging with fluorescently tagged YBX1
|
|
|
|
- Mass spectrometry to identify post-translational modifications
|
|
|
|
- Mutagenesis of key regulatory sites
|
|
|
|
- **Expected outcomes:** Understanding of how metabolic signals control YBX1 localization and function
|
|
|
|
|
|
|
|
## 6. Metabolic regulation of gene expression
|
|
|
|
- **Hypothesis:** Different nutrient environments reshape the epigenetic landscape
|
|
|
|
- **Approach:**
|
|
|
|
- Treat hepatocytes with various conditions:
|
|
|
|
- Insulin stimulation
|
|
|
|
- Beta-oxidation inhibitors
|
|
|
|
- Glycolysis inhibitors
|
|
|
|
- ATAC-seq to map chromatin accessibility changes
|
|
|
|
- RNA-seq to identify expression changes
|
|
|
|
- Metabolomics to correlate with cellular metabolic state
|
|
|
|
- **Expected outcomes:** Map how specific metabolic pathways influence gene regulation
|
|
|
|
|
|
|
|
## 7. Environmental effects on hepatocyte cell fate
|
|
|
|
- **Hypothesis:** Environmental factors reprogram hepatocytes through KLF-mediated mechanisms
|
|
|
|
- **Approach:**
|
|
|
|
- HepaRG differentiation under various conditions
|
|
|
|
- ChIP-seq for KLF family members
|
|
|
|
- CelSeq2 for single-cell trajectory analysis
|
|
|
|
- Functional validation of key target genes
|
|
|
|
|
|
|
|
## 8. Mice studies
|
|
|
|
- **Hypothesis:** Hepatocyte-specific YBX1 deletion protects against diet-induced fatty liver
|
|
|
|
- **Approach:**
|
|
|
|
- Generate hepatocyte-specific YBX1 knockout mice
|
|
|
|
- High-fat diet challenge
|
|
|
|
- Histological and biochemical analysis
|
|
|
|
- Multi-omic profiling of liver tissue
|
|
|
|
|
|
|
|
## 9. Miscellaneous projects
|
|
|
|
|
|
|
|
## 10. Spheroids/Organoids
|
|
|
|
- **Hypothesis:** 3D culture systems better recapitulate YBX1 function in vivo
|
|
|
|
- **Approach:**
|
|
|
|
- Establish liver spheroid/organoid cultures
|
|
|
|
- Manipulate YBX1 expression
|
|
|
|
- Single-cell RNA-seq for heterogeneity analysis
|
|
|
|
- Lipid loading experiments and imaging
|
|
|
|
|
|
|
|
## 11. Robot
|
|
|
|
- **Goal:** Automated high-throughput screening platform
|
|
|
|
- **Applications:**
|
|
|
|
- Drug screening for metabolic disease
|
|
|
|
- Systematic CRISPR screening
|
|
|
|
- Automated lipid accumulation assays
|
|
|
|
|
|
|
|
## 12. Directional RNAseq
|
|
|
|
- **Hypothesis:** Antisense transcription contributes to metabolic gene regulation
|
|
|
|
- **Approach:**
|
|
|
|
- Directional RNA-seq in normal vs lipid-exposed cells
|
|
|
|
- Integration with ChIP-seq data
|
|
|
|
- Functional validation of key antisense transcripts
|
|
|
|
|
|
|
|
## 13. DuoLink
|
|
|
|
- **Goal:** Map protein-protein interactions in situ
|
|
|
|
- **Applications:**
|
|
|
|
- YBX1-CEBP interactions in different cellular compartments
|
|
|
|
- Dynamic changes in protein complexes during lipid stress
|
|
|
|
|
|
|
|
## 14. FA Uptake
|
|
|
|
- **Hypothesis:** YBX1 regulates expression of fatty acid transporters
|
|
|
|
- **Approach:**
|
|
|
|
- Fluorescent fatty acid uptake assays in control vs YBX1-depleted cells
|
|
|
|
- ChIP-seq for YBX1 at fatty acid transporter gene loci
|
|
|
|
- Rescue experiments with transporter overexpression
|
|
|
|
|
|
|
|
## 15. Proliferation
|
|
|
|
- **Hypothesis:** YBX1 balances proliferation and differentiation in hepatocytes
|
|
|
|
- **Approach:**
|
|
|
|
- EdU incorporation assays
|
|
|
|
- Cell cycle analysis in YBX1-manipulated cells
|
|
|
|
- Integration with RNA-seq data
|
|
|
|
|
|
|
|
## 16. Lipidomics
|
|
|
|
- **Goal:** Comprehensive profiling of lipid species changes
|
|
|
|
- **Applications:**
|
|
|
|
- YBX1 knockout effects on hepatocyte lipid composition
|
|
|
|
- Temporal changes during fat-induced cellular reprogramming
|
|
|
|
|
|
|
|
## 17. Oxylipins
|
|
|
|
- **Hypothesis:** YBX1 regulates inflammatory signaling via oxylipin metabolism
|
|
|
|
- **Approach:**
|
|
|
|
- Targeted oxylipin profiling
|
|
|
|
- Expression analysis of oxylipin biosynthetic enzymes
|
|
|
|
- Functional validation with specific inhibitors
|
|
|
|
|
|
|
|
## 18. CelSeq time series
|
|
|
|
- **Hypothesis:** Fat exposure creates heterogeneous cell populations with distinct trajectories
|
|
|
|
- **Approach:**
|
|
|
|
- CelSeq2 time course during fat exposure
|
|
|
|
- Trajectory analysis and pseudotime ordering
|
|
|
|
- Identification of cell state markers
|
|
|
|
|
|
|
|
## 19. In vitro cytonuclear proteomics
|
|
|
|
- **Goal:** Map protein localization changes during metabolic stress
|
|
|
|
- **Approach:**
|
|
|
|
- Subcellular fractionation followed by mass spectrometry
|
|
|
|
- YBX1 interactome in different cellular compartments
|
|
|
|
- **Specific focus:** Epigenetic memory mechanisms and chromatin mismatch repair
|