Katia Koelle Sarah Cobey Bryan Grenfell Mercedes Pascual

1
Epochal evolution shapes the phylodynamics of
interpandemic influenza (H3N2)
Katia Koelle Sarah Cobey Bryan Grenfell
Mercedes Pascual
?
SI87
VI75
?
BK79
EN72
TX77
HK68
DIMACS, 9-10 October 2006
2
Pathogen diversity and cross-immunity
s
3
Modeling Cross-Immunity

• Strains with high sequence similarity must
  have high cross-immunity
• Strains with low sequence similarity must
  have low cross-immunity

4
Explaining limited diversity of hemagglutinin
Strain-specific cross-immunity
Actual HA1 phylogeny
Simulated phylogeny
Explosive diversity
Ferguson, Galvani, Bush, Nature (2003)
5
Explaining limited diversity
Strain-specific cross-immunity generalized
immunity
Limited diversity
Ferguson, Galvani, Bush, Nature (2003)
6
Modeling cross-immunity between flu strains

• Can sequence evolution be used as a proxy for
  antigenic evolution when modeling influenzas
  hemagglutinin?
• (i.e. does genotype approximate phenotype?)

• Propose alternative to this genotype-phenotype
  map for influenzas hemagglutinin evolution

• Consider the effect of this new mapping on the
  phylogenetics and dynamics (i.e. phylodynamics)
  of influenza H3N2

7
Influenza clusters
Cluster designations as in Smith et al. 2004
8
Topology of influenza clusters

• Strains with high sequence similarity can
  have low cross-immunity
• Strains with low sequence similarity can have
  almost complete cross-immunity

Genotype cannot serve as a proxy for antigenic
phenotype
9
STRAIN 1
STRAIN 2
Genotype-phenotype mapping?
10
Genotype-phenotype mapping for RNA 2o structures
Fontana Schuster, JTB (1998)
11
Neutral networks
Fontana Schuster, JTB (1998)
12
Evolutionary dynamics on neutral networks
Fontana Schuster, JTB (1998)

• A neutral mutation does not change the
  phenotype but it does change the potential for
  change What appears to be a sudden and abrupt
  change at the phenotypic level has been the
  result of neutral genetic drift. -Fontana

13
Neutral network mapping for proteins
Lau and Dill

• Single sequence changes can result in large
  changes in protein conformation.
• Changing a sequence by a large number of
  mutations may have no appreciable effect on
  protein conformation.

14
Implications for modeling cross-immunity
Bornberg-Bauer Chan, PNAS (1999)
Bornberg-Bauer
15
Modeling influenzas hemagglutinin
15 a.a. (45 nucs.)
5 epitopes
16
Changing the shape of an epitope

• Adaptation of Kauffmans NK model that
  generates neutral networks in genotype space
  (Newman and Engelhardt)

3

• Framework assumes epistatic or
  context-dependent interaction between amino acids
  located in the same epitope

15 a.a.
5 epitopes
17
Neutrality and sequence evolutionsubbasins,
portals, and epochal evolution
?
SI87
VI75
?
BK79
EN72
TX77
HK68
Adapted (for flu ? ) from Crutchfield, 2002
18
Coupling to an epidemiological model
Infected
Clusters
Adapted for clusters, from Gog Grenfell, PNAS
(2002)
19
Dynamic Consequences of Neutral Network Model
Years

• Cluster transitions
• Peaks in incidence during
• cluster transition years
• Refractory year

20
Comparison with observed influenza dynamics
Greene et al. (2006)
21
Phylogenetic Consequences
Simulated tree
Observed HA tree (from Smith et al. sequences)

• Explosion of diversity within clusters
• Cluster transitions cause selective sweeps
• No need for generalized immunity to limit HA
  diversity

22
Expected pattern in genetic diversity arising
from epochal evolution
23
Supporting empirical evidence
24
Notions of neutrality
Influential sites model Only changes at very few
sites can precipitate a cluster jump, and their
ability to do so does not depend on the genetic
background in which they occur. Genetic
diversification within clusters does not
facilitate adaptive change, and can be safely
ignored. Context-dependent model Changes at most
sites can precipitate a cluster jump if those
changes occur in the right genetic
background. Cluster innovations are guided by
the process of neutral diffusion, via changing
the genetic background of sequences.
See also Wagner, 2005 for a discussion on types
of neutrality in non-flu systems
25
Importance of genetic background,
i.e. context- dependency
Influential sites
26
Observed pattern in genetic diversity
Boom-and-bust of genetic diversity empirically
supported
27
Observations of tree balance
Diversification within clusters cannot be
rejected under the null, neutral model of random
speciation.
28
Conclusions

• An alternative, empirically-supported model of
  influenzas hemagglutinin evolution can account
  for both H3N2s dynamic and the phylogenetic
  patterns of its HA1.
• Incorporating appropriate genotype-phenotype maps
  for the effect of mutations at the phenotypic
  level may be important for understanding pathogen
  evolution.

29
Acknowledgments
David Alonso, Stefano Allesina, Luis Chaves,
Diego Moreno, Aaron King Center for the Study of
Complex Systems
NSF graduate student fellowship (S.C.) McDonnell
Foundation (Centennial Fellowship to M.P.)
Jamie Lloyd-Smith, Igor Volkov, Mary Poss CIDD
postdoctoral fellowship (K.K.)
Derek Smith, Ron Fouchier, Sharon Greene, Cecile
Viboud, Maciej Boni
30
(No Transcript)
31
(No Transcript)
32
Patterns of influenza phylodynamics (H3N2)
1. Annual outbreaks
Greene et al. (2006)
3. Genetic change
Antigenic change
2. Genetic drift
Fitch et al. (1997)
Smith et al. (2004)
33
Patterns of genetic diversity
34
Characteristics of Influenza Evolution
Sequential replacement of clusters
Cluster
Season
Smith et al., Science (2004)
35
Characteristics of Influenza Evolution
Genetic distance from 1968 strain
Antigenic distance from 1968 strain
Smith et al., Science (2004)