A network perspective on the evolution of aging

1
A network perspective on the evolution of aging

• Daniel Promislow
• University of Georgia

2
Two basic questions

• Why do we age?
• Why do some individuals in a population age
  faster than others?

3
Two basic answers

• Proximate (or how) explanations
• Reactive oxygen species ? oxidative damage
• Telomere shortening ? cell cycle arrest
• Repair mechanisms fail ? DNA damage
• Ultimate (or why) explanation
• Force of selection declines with age ? rates of
  survival and fecundity decline

4
Life cycle graph
bn
b2
b1
2
3
n
1
P1
P2
Pn
P age-specific survival rate F age-specific
fecundity
5
Life cycle graph
bn
b2
b1
2
3
n
1
P1
P2
Pn
Mutation Accumulation (Medawar)
6
The challenge
Physiology, genetics, biochemistry, etc.
Evolutionary Theory
7
Networks
Node
8
(No Transcript)
9
Degree Distribution for Amazon.com purchasing
preference network
Number of observations
Connectivity (number of pointers)
Data compiled by Eric Promislow (2003)
10
Yeast protein-protein interactions
11
Networks and the Biology of Aging

• Description
• Do aging genes have unique network structure?
• What do aging networks look like?
• Prediction
• Can we use networks to predict genes or functions
  related to aging?
• Theory and its applications
• How might selection prevent or fail to prevent
  aging?

12
Yeast protein-protein interactions
13
Connectivity and aging genes

• We assume that more highly connected genes are
  more pleiotropic
• More highly pleiotropic genes more likely to
  exhibit antagonistic pleiotropy
• More highly connected genes more likely to be
  associated with aging

14
Connectivity and pleiotropy
Promislow, 2004. Proc. Royal Society
15
Connectivity and aging
Promislow, 2004. Proc. Royal Society
16
Connectivity and aging
Can hubs explain natural variation in aging?
The one hand No, hubs under strong selection for
robustness. The other hand Yes, minor
changes in hub function will propagate.
17
Aging and robustness

• Highly connected genes more likely to affect
  lifespan
• Large body of theory on networks and robustness,
  but not yet applied to aging
• Need to understand how selection acts, or fails
  to act, on network components

18
Description 2.0
Molecular Systems Biology 2007
19
(No Transcript)
20
Aging in dogs
21
VMDB database

• 184 AKC breeds
• 7857 Diagnostic codes
• 80,973 deaths 1984 - 2004

22
Aging and Epidemiology
Comorbidity 1 1 4 3 1
23
Frequency distribution of disease states
(comorbidity)
Exponential distribution
24
Frequency distribution of disease states
(comorbidity)
R20.941
R20.996
25
Implications of power law distribution

• Power law fits better than the exponential
• Network distributions approximate power law
• probability of new connections ? with node degree
• Among dogs, the sick get sicker
• Network node accumulation analogous to
  age-related disease accumulation?

26
Rzhetsky et al. 2007. PNAS
27
Network plasticity robustness

• Plasticity response to environmental ?
• Robustness inverse of plasticity
• Two kinds of robustness
• Environmental
• Constancy in the face of environmental
  perturbations
• Genetic
• Constance in the face of mutations

28
Gene regulatory networks
29

• Gene expression varies across environments
• Is plasticity (or its inverse--robustness)
  related to network structure?

Gasch et al., Mol. Biol. Cell 2000
30
More plastic genes have more regulators
R2 9.5
Plasticity
log(Connectivity)
Promislow, American Naturalist 2005
31
Selection Environmental Plasticity
Selection is strongest in highly
plastic genes highly regulated genes
Robustness may not always be a good thing
Environmental Plasticity
log(Ka/Ks )
32
Selection Environmental Robustness
Environmental Robustness
log(Ka/Ks)
strong selection
weak selection
33
Selection and Genetic Robustness
Selection is strongest in highly genetically
robust genes
Genetic Robustness
log(Ka/Ks )
strong selection
weak selection
34
A paradox?
ER
GR

• Selection favors strong environmental plasticity,
  but strong genetic robustness Antagonistic
  Pleiotropy?
• Similar mechanisms may underlie ER and GR
  (transcription factors? heat shock proteins?)
• These mechanisms may constrain GR over the
  lifetime

35
Implications for Aging

• Plasticity and Robustness
• common mechanisms, conflicting selection pressure
• Selection is strongest early in life
• Effects of deleterious mutations accumulate with
  age
• Early-life selection favors plasticity, at a cost
  of decreased robustness late in life
• If we shift selection to late age, we should see
  less plasticity. This could explain why
  short-lived flies are more responsive to
  longevity-inducing factors

36
Networks and the Biology of Aging

• Description
• Do aging genes have unique network structure?
• What do aging networks look like?
• Prediction
• Can we use networks to predict genes or functions
  related to aging?
• Theory and its applications
• How might selection prevent or fail to prevent
  aging?

37
Conservation of aging-related genes
Observed Frequency of Orthologs with Conserved
Role in Aging
?
Really 3.5 if only yeast genes with worm
orthologs are considered
38
276 Worm Longevity Genes
Baseline 3.5
39
Predictive networks

• Generate shortest-path network of genes
  associated with aging in yeast
• Test aging phenotypes in connected nodes

40
Linked nodes are enriched for life-extension in
knock-outs
Both Haploid Mating Types
41
Networks and the Biology of Aging

• Description
• Do aging genes have unique network structure?
• What do aging networks look like?
• Prediction
• Can we use networks to predict genes or functions
  related to aging?
• Theory and its applications
• How might selection prevent or fail to prevent
  aging?

42
Rate of progression of cell lineages towards
cancer
43
Robustness, cancer aging

• Cancer arises when n components fail
• As n increases, more robust against cancer
• But, each component is more likely to
  accumulation deleterious mutations
• More components leads to more robustness, but
  greater genetic variation
• Can eventually lead to increased rates of cancer
• May provide a useful heuristic model of aging

44
Conclusions

• Network-based systems approaches
• A shift away from one-gene ? one-trait thinking
• Epistasis rules
• Genes related to aging may have unique network
  structure
• Networks may help us to predict genes or
  functions related to aging.
• Need for more theory
• why are some traits more robust than others as we
  age?
• What maintains genetic variation for aging?

45
Acknowledgements

• University of Georgia
• Jake Moorad
• Chrissy Spencer
• University of Washington
• Brian Kennedy
• Matt Kaeberlein
• Erica Smith
• U.C. Santa Barbara
• Steve Proulx
• UTHSC San Antonio
• Steve Austad
• Ellison Medical Foundation
• NSF, NIH