Changes in Protein Abundance Across Nutrient

Changes in Protein Abundance Across Nutrient Conditions
Sean Hackett, Jonathan Goya, David Perlman and Josh Rabinowitz
Across varying nutrient conditions, with differing limiting nutrients and maximal
growth rates, cells must deal with two challenges:
① Increasing fluxes (rates of enzymatic reactions) in parallel with nutrient
② Sensing which nutrients are limiting and growing optimally with respect to that
To study this control of metabolism, previous researchers in the Botstein and
Rabinowitz labs have measured transcript and metabolite abundance across
25-36 steady state conditions.
Limiting nutrients: glucose (C), ammonium (N), phosphate (P), uracil (U) and
leucine (L) (in auxotrophs).
Dilution rates: 0.05, 0.11, 0.16, 0.22, 0.30 h-1
Going from peptides to proteins
Working with peptides
Of the 11434 peptides that were measured in at least one of the
15 conditions, the coverage of peptides that are quantifiable
under all 15 conditions is quite poor
We need methods that are:
• Able to use peptides with missing data to understand patterns
in complete data
• Look at patterns of missing data to identify peptides that are
present in only a subset of conditions
Number of conditions where
a peptide is quantifiable
Overlap of peptides measured in both
conditions relative to just one
• Remove all peptides with more than 0.2 fraction of missing data
• Remove all peptides with less than 32768 ions measured
• This results in 5106 peptides that are retained and 6328 peptides which are discarded.
When using relative variation in peptides to predict variation in protein abundance we need
to deal with two factors
① If a peptide maps to multiple proteins, it should be attributed to a protein (thereby adding
signal) to the extent that its pattern matches the pattern from other peptides.
② Some peptides won't conform to the trends of their protein because they may be the noncovalently modified complement of a set of unascertained modified peptides. These
peptides shouldn't inform the general protein trend, and may be interesting to analyze in
Examples of fits
Comparing transcripts and proteins
Raw Ω
Impute missing values with
PCA-based method
Brauer 2008
Boer 2010
In a followup to these studies, we are combining these metabolic and transcriptional
datasets with quantitative proteomics, rates of nutrient uptake and secretion and
composition measurements.
Experimental protocol:
- Grow FY4 S. cerevisiae to steady state under the desired nutrient limitation and
dilution rate
- Filter culture and freeze on li-N2
- Homogenize and extract proteins
- Combine sample proteins 1-1 with 15N labeled reference chemostat (p0.05)
- Digest combined proteins with trypsin
- Fractionate peptides into 24 fractions using isoelectric focusing
- Identify and quantify peptides using time of flight tandem mass spectrometry on an
Agilent Q-TOF
Computational (by JG):
- MS1 data (peaks in 2D: elution time * mass-to-charge) gives total ions, with
uncertainty about the identity of the ion
- Compare MS2 data (peptide sequences) to predicted peptides based on protein
sequences, passing identity to some MS1 peaks
- Align MS1 data across files to pass identities between peaks identified in only a
subset of conditions
From brauer 2008 infer relative abundances at
dilution rates attained in this study
How correlated are proteins and
their corresponding transcripts ?
Future Directions
• What types of proteins are present or absence at
high and low growth rates, or are limitation-specific
• What types show a quantitative relationship to
growth rate or a limitation-specific effect
• What is responsible for the poor correlation of
many proteins vs. transcripts
• How much can these patterns be explained by
motifs in DNA, RNA or proteins
• Modeling of fluxes and studying metabolic versus
hierarchical control and the sufficiency of
established reaction mechanisms

similar documents