Protein-lipid interactions in influenza virus entry Peter Kasson - - PowerPoint PPT Presentation

SLIDE 1

Protein-lipid interactions in influenza virus entry

Peter Kasson Departments of Molecular Physiology and Biomedical Engineering University of Virginia

Tuesday, May 12, 15

SLIDE 2

Why is it hard to predict pandemics

0+ 100+ 200+ 300+ 400+ 500+ 600+ 700+ 800+ 900+ 1000+ 1995+ 2000+ 2005+ 2010+ 2015+

Cases Sources: US CDC; Russell, Kasson et al., 2014 Year

Tuesday, May 12, 15

SLIDE 3

Cell entry by influenza

Viral membrane Target cell

Fusion

Target and physical environment for fusion

mutations block

FIGURE 8

Stages of fowl pl ague vi rus ent ry into MDCK cel ls . Ce l ls wi th prebound vi rus were warmed at 37°C for di f ferent t imes

and then f ixed wi th glutara ldehyde at room temperature . Wi thin 5 min, vi rus par t ic les were seen

in smooth sur faced pi ts and

vesic les (a, b, and c) , coated pi ts (d, e, and f) and coated vesic les (g, h, and

i )

. I n

f, the samp l e was sta ined wi th ant i - fowl pl ague

vi rus spi ke prote in IgG and then wi th ferr i t in-goat ant i - rabbi t IgG af ter forma ldehyde f ixat ion (see Mater i a l s and Methods) . This

image demonst rates that par t of the vi rus par t icle was t ight ly assoc i ated wi th the membrane since onl y the exposed par t

is l abe l ed

wi th ferr i t in . Af ter 10 min, vi ruses were observed in endosomes ( j ) and mul t ivesi cul ar bodi es (k and I) . The images shown in

a, b,

and c were af ter 2 min warming ; in d, e, g,

k, and

i af ter 5 mi n warming, in f af ter 1 min warming, and in j , k, and / af ter 10 min

warming . a- i , x 62,500; k- l , x 50,000 .

brane of cel ls by lower ing the medium pH (5, 19, 54) . I f fowl plague vi rus infects MDCK cel ls by an endocytot ic pathway passing through the lysosomes,

i t might also be expected to

fuse at the plasma membrane i f exposed to low pH . This

seemed especial ly l ikely since low pH-dependent hemolysis and cel l -cel l fusion had been recent ly demonst rated for inf lu- enza vi ruses (19, 55, 56) . To test this, cel ls wi th prebound vi rus were suspended in media of pH 5 .0 and pH 7.4 for

1 min at

37°C and examined by t ransmission elect ron microscopy af ter

indi rect ferr i t in immunolabe l ing. In the cel ls kept at pH 7 .4, ferr i t in was associated only wi th vi rus part icles and not wi th the cel l sur face . No fusion of the vi rus wi th the cel l sur face was

• bserved.

In contrast ,

ferr i t in was at tached to the plasma

membrane only in samples exposed to low pH (Fig . 10) . In

several cases, clear cont inui ty between the cel l and vi rus mem-

branes was observed, wi th ferr i t in only bound to the prot ruding

vi rus prof i le (Fig . l0 a and b) . Fusion of vi ruses to membrane vesicles apparent ly shed f rom the cel ls was also observed (not

shown) . Low pH t reatment of MDCK cel ls in the absence of

vi rus produced some disturbance of the plasma membrane but did not induce art i factual ferr i t in binding .

To quant i tate low pH- induced fusion of fowl plague vi rus

to the MDCK cel l plasma membrane, cel ls wi th prebound radioact ive vi rus were incubated for 30 s at 37°C wi th media

FIGURE 10

Fus ion of fowl pl ague vi rus at the MDCK pl asma membrane . Fowl pl ague vi rus (60 j .g) was bound to MDCK cel ls for 1 h at 0°C and fusion was induced by incubat ing the cel ls for

1 min at pH 5 .0 and 37 °C. The cel ls were

then f ixed wi th forma ldehyde and immuno l abe l ed wi th ant i - fowl pl ague vi rus spi ke IgG and fer r i t in-conjugated goat ant i - rabbi t IgG . Vi ruses are

c lear ly recogni zabl e by the ferr i t in at tached to the i r membrane . In a the nuc l eocaps id

is st i l l c lear ly visible . In b the caps id is bare ly

recogni zabl e, but the vi rus shape and ferr i t in- labe led spi kes are st i l l obv ious

. In c the spi ke prote ins have presumabl y di f fused in

the pl ane of the membrane away f rom the si te of fusion . A vi rus prof i le

is st i l l detectabl e (ar row)

. Bar 0 .2 j m . x 106,000 .

MATU i v

FT At .

Inf luenza Vi rus Ent ry

609

FIGURE 8

Stages of fowl pl ague vi rus ent ry into MDCK cel ls . Ce l ls wi th prebound vi rus were warmed at 37°C for di f ferent t imes

and then f ixed wi th glutara ldehyde at room temperature . Wi thin 5 min, vi rus par t ic les were seen

in smooth sur faced pi ts and

vesic les (a, b, and c) , coated pi ts (d, e, and f) and coated vesic les (g, h, and

i )

. I n

f, the sampl e was sta ined wi th ant i - fowl pl ague

vi rus spi ke prote in IgG and then wi th ferr i t in-goat ant i - rabbi t IgG af ter forma ldehyde f ixat ion (see Mater i a ls and Methods) . This

image demonst rates that par t of the vi rus par t icle was t ight ly assoc i ated wi th the membrane since onl y the exposed par t

is l abe l ed

wi th ferr i t in . Af ter 10 min, vi ruses were observed in endosomes ( j ) and mul t ivesi cul ar bodi es (k and I) . The images shown in

a, b,

and c were af ter 2 min warming ; in d, e, g,

k, and

i af ter 5 min warming, in f af ter 1 min warming, and in j , k, and / af ter 10 min

warming . a- i , x 62,500; k- l , x 50,000 .

Matlin et al., 1981

Tuesday, May 12, 15

SLIDE 4

Influenza fusion is heterogeneous on a single-virus level

fusion efficiency 10-40% depending on conditions in-cell fusion efficiency ~10% even if we could simulate relevant timescales, a single movie wouldn’t do it

Waiting Time (s)

50 100 150 200

Num Fusion Events

5 10 15

Single fusion events detected via fluorescence dequenching

Tuesday, May 12, 15

SLIDE 5

Bad combination

Heterogeneous outcomes -> need many simulations for statistics Slow decorrelation times -> need long simulations Simplest full-scale systems >>1M particles -> need large simulations Biological system sensitive to fine details -> need high-fidelity simulations

Tuesday, May 12, 15

SLIDE 6

Unraveling virus-membrane interactions surrounding fusion

What are the relevant physical interactions controlling influenza viral fusion? Building an integrated understanding from simulations and biophysical experiments. Today: membrane interfaces preceding fusion, membrane-protein interactions.

Tuesday, May 12, 15

SLIDE 7

Multi-pronged approach

“mid-scale” systems ~1-3M atoms isolated components 150K-800K atoms

Building integrated understanding via statistical models at multiple levels

Tuesday, May 12, 15

SLIDE 8

Membranes form stable interfaces prior to fusion

Unexpected! Now good indirect experimental evidence!

Depending on the system, these can be 10’s of ns to ~10 μs

Tuesday, May 12, 15

SLIDE 9

Decreased water mobility at vesicle interface

JACS 2011

Tuesday, May 12, 15

SLIDE 10

Glassy dynamics of water between two lipid membranes

Pronk, Lindahl, Kasson. JACS 2015

5 10 15

Nw/Nl

5 10 15 20

<tp> / <tx>

with lipid diffusion correction without correction lipids single bilayer

Implications for simulating fusion dynamics--can get stuck!

Tuesday, May 12, 15

SLIDE 11

Multi-level parallelism

Need both parallelism at the individual “partition” level (MD scaling over N cores/GPU’s) and parallelism between partitions in solving the overall statistical problem

SIMD SIMD SIMD SIMD SIMD SIMD SIMD SIMD SIMD SIMD SIMD SIMD

Thread Thread Thread MPI MPI MPI Worker Worker Worker Server

IB SSL Shared memory

Average: 0.04MB/s Peak: 100MB/s Latency: 10 ms Average: 0.5GB/s Peak: >2.7GB/s Latency: 1-10μs Average: 0.5GB/s Peak: 25GB/s Latency: <100ns Cluster

Pronk et al., 2011; Pronk et al., 2015

Initial sampling model generation

parallel simulations

Currently doing this ad-hoc on BW Copernicus: DAG engine to coordinate this

generate next round

Tuesday, May 12, 15

SLIDE 12

Hemagglutinin transmembrane domains

• Hemagglutinin is trimeric; multiple trimers likely act together in fusion
• Truncations to the TM domain can arrest fusion
• Could TM-TM interactions play a role in fusion?
• Multi-resolution approach to characterize TM interactions
• Use coarse graining to sample diffusional processes, atomistic simulations to

sample conformational equilibria

many simulations many simulations many starting conditions many conformational samples statistical characterization

Tuesday, May 12, 15

SLIDE 13

Reproducible, stable trimerization by TM domains

Residue-Residue Contact Probability

181 186 191 196 201 206 211 216 181 186 191 196 201 206 211 216

0.1 0.2 0.3 0.4 0.5 0.6

Contacts primarily “in register”

Probability map of inter-monomer contacts from 50 atomic-resolution simulations

residue j residue i

time (us)

2 4 6 8 10 12

distance (nm)

0.5 1 1.5 2 2.5 3 3.5 4

Peptide-Peptide Center of Mass Distance : CG

Coarse-grained simulations form TM trimers on the ~2 μs timescale

All 24 simulations formed trimers

Tuesday, May 12, 15

SLIDE 14

TM domain trimers are robust to mutation

Mutating the top 4 contacts abrogates those contacts but does not disrupt trimer. This is consistent with experimental mutational data

time (ns)

50 100 150

distance (nm)

0.5 1 1.5 2 2.5 3 3.5 4 4.5

Peptide-Peptide Center of Mass Distance : Quadruple Mutant

residue j residue i change in inter-monomer contacts average over 50 simulations

Tuesday, May 12, 15

SLIDE 15

Building integrated understanding...

• Methods: integrate statistical models into high-level parallelism
• Statistical models of protein-membrane dynamics for different interactions

involved in influenza viral entry

• Statistical models of influenza-mediated membrane fusion
• Integrating with biophysical experiments

Ultimately, all of this is a single large sampling and statistical learning problem. Hard because we don’t know the relevant reaction coordinates.

Residue-Residue Contact Probability 181 186 191 196 201 206 211 216 181 186 191 196 201 206 211 216 0.1 0.2 0.3 0.4 0.5 0.6

Waiting Time (s) 50 100 150 200 Num Fusion Events 5 10 15

Tuesday, May 12, 15

SLIDE 16

Acknowledgements

Group members

• Bert Cortina
• Matt Eckler
• Jennifer Hays
• Per Larsson
• Malgorzata

Latallo

• Bob Rawle
• Rich Salaway
• Jeff Xing
• Katarzyna

Zawada NIGMS R01 GM098304 The Hartwell Foundation STINT

Funding & Resources:

Collaborators

• Erik Lindahl (SciLife)
• Sander Pronk
• Vijay Pande (Stanford)
• David Steinhauer

(Emory)

• Kelly Dryden
• Lukas Tamm
• Sonia Gregory
• Judith White
• Linda Columbus
• Steven Boxer

(Stanford)

• Mark Yeager

Tuesday, May 12, 15