01-Title [Converted]

Three-Dimensional Organization of Chromosome Territories and the Human Cell Nucleus

? Tobias A. Knoch, Christian Münkel, Jörg Langowski Biophysics of Macromolecules

German Cancer Research Center (DKFZ) Heidelberg - Germany

The dynamic and hierarchical organization of cell nuclei span between 10 and 13 orders of magnitude concerning length and time scales.

base pair

chromosome (metaphase)

chromatin loop

B-DNA superhelix

chromosome band

Human Human 3DGenome Genome Study Group

Tobias A. Knoch

nucleus (interphase)

DNA length 100

Volume Time

-9 10-8

10-10

101

102

10-7

10-8

103

10-5

10-6

104

105

106

10-3

10-4

107

108

100

10-2

100

109

1010

bp

103

m 3

103

s

Overview

Human 3D Genome Study Group

Tobias A. Knoch

Experiment

Simulation

Genomic Region (Chromosome or Gene)

Multi-LoopSubcompartment and Random Walk/ Giant Loop model

fluorescence in-situ hybridization (FISH)

3D confocal scanning microscopy

polymer model for simulation of the chromatin fiber

Conclusions for the human cell nucleus chromosome-, chromosome-arm and subcompartment overlap 3D-distances between genomic markers as function of their genomic separation behaviour of marker ensembles and dynamics of structural features fractal properties of chromosomes decondensation of chromosomes from metaphase into interphase and chromosome stretching conclusions from simulating whole cell nuclei

Fluorescence in-situ Hybridization

Human 3D Genome Study Group

FISH

Tobias A. Knoch

Cell - Preparation

Probe - Preparation

cells on coverslip grown to confluent layer

finding of genomic site for marking and cloning of this sequence AACGG TTGCC

fixation of cells on coverslip (formaldehyde) and permeabilisation

labeling of the DNA probe (Nick translation or PCR) with Digoxigenin Fluorophor (indirect) (direct) AACGG TTGCC

DNA double strand TACGTTAACGGTAGCATT ATGCAATTGCCATCGTAA

AACGG TTGCC

Hybridization

probe is put on coverslip and melting of the double strands at 70C

TA A

CG

GT AG

CA TT

TTGCC

CC

TA C

TTGCC

TA C

GT

TTG

GG AACGTAA TC CCA

GT

TTG CAA

ATG

amplification with fluorescent labeled antibodies TACGTTAACGGTAGCATT TTGCC

TA A TT CGG GC TA C GC

AT T

Chromosomes form distinct territories in interphase and genomic markers lie within the territories and are clearly separable. Left: Territory painting by FISH of chromosome 15; by chance the two territories neighbour each other. Right: Genomic markers YAC48 and YAC60, genomic separation 1 Mbp.

Human Human 3DGenome Genome Study Group

Tobias A. Knoch

Ideogram of chromosome 15 with Prader - Willi Region and Angelmann Region. The size and genomic distance of the clones are sufficiently small and well characterized to measure the fine structure and organization of chromosome territories.

Human 3D Genome Study Group

Tobias A. Knoch

Prader-Willi Region (paternal chromosome) YAC 48 Angelmann Region (maternal chromosome) YAC 60

arm

1Mbp

centromer clones and their genomic size

λ 48.1 p

λ 48.17

17,3

arm

17,8 69.2

[kbp]

λ 48.14

11.0

13.0 125,7

λ 48.7

19,0

194.9 144.7 213.9

genomic distances between probe pairs

q

Dual colour FISH of genomic markers leads to measurements of 3D-distances which are below the resolution of the microscope. Critical signals could also be excluded with higher confidence. Genomic marker λ48.1 in red and marker λ48.14 in green, genomic separation 195 kbp.

Study Group

Tobias A. Knoch

[ d

[ [

dual colour

[

one colour

Human Human 3DGenome Genome

d?

d

d

Statistical analysis of the spatial distances between the PWS-Region (YAC48) and AS-Region (YAC60) with a genomic distance of 1Mbp = 10m chromatin fiber.

Human 3D Genome Study Group

Tobias A. Knoch

Distance Distribution

smooth fit normal fit

5

Spatial Distance [nm]

2600

2400

2200

2000

1800

1600

1400

1200

1000

800

[N] 600

[%]

10

400

.03

15

200

.05

20

0

Frequency

.08

# of distances: 305 Mean: (77623)nm Shapiro Wilk test on normality: W=0.87

The Multi-Loop-Subcompartment (MLS) and the Random Walk / Giant Loop (RW/GL) Model. Rosettes in the MLS-Model correspond to the size of chromosomal interphase band domains.

Random Walk / Giant Loop model (RW/GL) Sachs et al. (1995)

Study Group

Tobias A. Knoch

Multi-Loop-Subcompartment model (MLS) Münkel et al. (1997)

R bands

G bands

loop size: 3Mbp-5Mbp

backbone (non DNA)

Human 3D Genome

loop size: 126kbp

attachment points chromatin fiber, 30nm thick

rosette size: 1-2Mbp (according to interphase ideogram bands)

Linker consists of DNA (126 kbp) (in contrast to backbone)

Polymer Chain and Potentails

The chromosome fiber is simulated assuming a polymer chain and harmonic potentials. No hydrodynamic interaction is used due to hydrodynamic shielding.

Human 3D Genome Study Group

Tobias A. Knoch

R(i+1)

Excluded Volume Potential Segment Length l

Bending Angle ß

R(i)

R(i+2)

Stretching Potential Us ( l ) =

k BT 2

2

Bending Potential U b( ß ) =

( l - l 0)

Excluded Volume Potential Uev( r ) = k

Uev0 k BT

B : Bolzmann constant

T : : : rc : Lk :

(

)

4 2 2 r 2r r c 1+ r c4

Temperature, 310 K stretching elasticity, = 0.1 bending elasticity minimum distance of segments Kuhn length, 300 nm, Lk = b0/2 2

k BT 2

2

ß

2

VirtNucSim

Human 3D Genome

The programme code is written in C++ and uses Message Passing Interface (MPI) for parallelization and scales well at least up to 64 processors depending on compilation.

Study Group

Tobias A. Knoch

VirtNucSim

Chrom Create CompNuc Geometry VirtMic

SimController Parameter

Configuration CellsGrid

Monte Carlo Brownian Dynamics Diffusion

Brownian Dynamics Steps per Second [N/s]

1.200

1.000

Iterator

Communicate

Load Balancing

Beads

VirtNuc - Brownian Dynamics: Scalability Pentium Pro 200 Mhz - IWR IBM SP2 R6000 66 Mhz- DKFZ Convex S Class 1600 paRisc1.1 120 Mhz- DKFZ Cray T3E 450 Mhz- HLRS

0.800

0.600

0.400

0.200

0.000

1

2

4

8

Processors

16

32

Linked-Cell Algorithm and Dynamic Load Balancing

A linked-cell algorithm reduces the computation time for the pairwise Excluded Volume interaction using all beads within one cell and half of its 26 neighbour cells. Dynamic Load Balancing reduces the computation time by projecting the 3D cell grid dynamically on the 2D processor grid (spherical nucleus time reduction: 1/3). To avoid communication overhangs asynchronious, buffered communication is used.

Human 3D Genome Study Group

Tobias A. Knoch

Segments/Beads

DLB

3D Cell Grid

(size dynamic, much finer than shown)

2D Processor Grid (dynamic)

Random-Walk/Giant-Loop model versus Multi-Loop-Subcompartment model. Simulation results of chromosome 15. The chromosome is simulated assuming a flexible polymer chain, starting with ~ 3500 segments of 300nm = 31kbp and relaxing with ~ 21,000 segments 50nm = 5.2kbp. The starting configuration has the approximate form and size of a metaphase chromosome.

Ray traced image of the Random-Walk/Giant-Loop model, loop size 5Mbp, after ~80.000 Monte-Carlo and 1000 relaxing Brownian-Dynamics steps. Large loops intermingle freely thus forming no distinct features like in MLS model.

Human 3D Genome Study Group

Tobias A. Knoch

Ray traced image of the Multi-Loop-Subcompartment model, loop size 126kbp, linker size 126 kbp, after~50.000 Monte-Carlo and 1000 relaxing Brownian-Dynamics steps. Here rosettes form subcompartments as separated organizational and dynamic entities.

Wire frame image of the starting configuration.

Chromosome arms and bands do not overlap. The MLS-model predicts this behavior.

Study Group

Tobias A. Knoch

Arm Overlap

0.2

Confocal images of interphase p- and q- arms of human chromosome 3

Chr 6 p/q overlap Chr 3 p/q overlap

0.15

frequency

Human 3D Genome

RW/GL

0.1

MLS

experiment

reconstruction

S. Dietzel

R. Eils

0.05 0

0

20 30 10 overlap volume [%]

40

Subcompartment Overlap

frequency

0.2

experiment

Confocal images of interphase R- and G- bands of human chromosome 15

RW/GL MLS

0.1

0

experiment

0

20 40 overlap volume [%]

60

D. Zink

simulation

Comparison of the RW/GL- and MLS model with experimentally determined interphase distances.

Best agreement between simulations and experiments is reached for a MLS-model with a loop size of 126kbp and a linker length of 126kbp.

mean spatial distance [m]

1.5

Fibroblasts 15q11-13 PLW/A-Region, FISH, our data. Lymphocytes 11q13 FISH, K. Monier, Institut } Albert Bonniot, Grenoble. Fibroblasts 11q13

Study Group

Tobias A. Knoch

RW/GL-model: loop size: 5 Mbp

3 Mbp 1.0 Mbp 0.5 Mbp

Fibroblasts 4p16.3 FISH, Yokota et. al., 1995.

1

Human 3D Genome

0.5 MLS-model (loop size 126 kbp): linker length: 251kbp

0

188kbp 126kbp 62kbp

0

0.2

0.4

0.6

0.8

genomic distance between genomic markers [Mbp]

1

Shift of a marker ensemble through a rosette in the MLS-model in respect to loop bases. This leads to different sets of 3D-distances for every ensemble position. Due to the symmetry of the MLS-rosettes periodicities are found.

Human 3D Genome Study Group

Tobias A. Knoch

Spatial Distance [m]

0.300

0.250

0.200

0.150

0.100 0

5

10

15

20

25

30

35

40

45

Marker Shift in Respect to Loop Base [*60kbp]

Genomic Marker Distance:

31kbp

145kbp

171kbp

215kbp

Log(Curve Length in Number of Measures)

In agreement with porous network research fractal analysis show multifractal behaviour in simulations of chromosome 15. Different fractal dimensions mean different process-dynamics in these spaces. Therefore chromosomal territories show a higher degree of determinism than previously assumed.

5

Euclidean Cut Off D = 1.1

Tobias A. Knoch

D = 2.43 D = 1.89 D = 2.48

2

D = 2.88

D = 2.93

D = 2.01

1

0 1.5

Study Group

RW/GL: Excluded Volume 0.1kT LoopSize 5Mbp, LinkerLength 3600nm, 20 Loops LoopSize 126kbp, LinkerLength 600nm, 561 Loops MLS: LoopSize 126kbp, Excluded Volume 0.1kT LinkerLength 600nm LinkerLength 1200nm LinkerLength 1800nm LinkerLength 2400nm

4

3

Human Human 3DGenome Genome

D = 2.98

D = 2.03

SegmentLength 300nm

2

2.5

3

Log (Measure Length [nm])

3.5

4

Creation of a ´Virtual Human Cell Nucleus´ with all 46 chromosomes using the MLS-model.

a) 46 metaphase configurations are randomly placed in a spherical potential and dencondensed into interphase by Brownian Dynamic or Monte Carlo methods. b) 46 chains of spheres (number of spheres ~ chromosome size) are randomly placed in a spherical potential and relaxed with Simulated Annealing. Then the fine structure is added to receive the same resolution as in a).

0 ms

10 ms

Human 3D Genome Study Group

Tobias A. Knoch

50 ms

´Virtual Human Cell Nucleus´ Simulation of all 46 chromosomes using the MLS-model.

The nucleus is simulated assuming a flexible polymer chain, modelling the 46 chromatin fibers with a total 1,248,794 segments of 50 nm = 5.2 kbp. Pictures are shown after a 0.5 ms Brownian Dynamics simulation, one step taking 10s.

3-D rendering

Human 3D Genome Study Group

Tobias A. Knoch

simulated confocal section

The MLS-model leads to low overlap of chromosome-arms and subcompartments in contrast to the RWGL-model. This is also seen in experiments.

Human 3D Genome Study Group

Tobias A. Knoch

nucleus 6 m diametre

nucleus 8 m diametre

nucleus 10 m diametre

nucleus 12 m diametre

Mapping of Histone H2B-GFP and H1-GFP distribution in vivo. The Histone-GFP reflects the distribution of chromatin in interphase. The structure visible in the images is similar to those found in simulations. HeLa cells stably transfected with H2B-GFP (K. Sullivan, Scripps Institute). Confocal in vivo section of a cell nucleus and a mitosis. Right: Cos7 cell stabaly transfected with H1-GFP (A. Alonso, DKFZ). Confocal in vivo section.

Human 3D Genome Study Group

Left:

1m

1m

Tobias A. Knoch

DNA fragmentation by irradiation with carbon ions. Irradiation with carbon ions results in DNA double strand breackage. The length of the fragments follow distributions depending o the spatial arrangement of the 30 nm chromatin fiber in the nucleus. Together with P. Quicken, GSF, Munich.

Radiation dose distribution in a cell nucleus irradiated with carbon ions

0.008

6

100000 10000 1000 100 10

4

osit

0

ion

-2 2

[m]

-4 4

yP

-2

os iti

0

6

on

[m ]

2

x-P

Tobias A. Knoch

MLS with 0.126 Mbp loops and linkers RWSL with 0.126 Mbp loops RWSL with 0.126 Mbp loops RWGL with 5.0 Mbp loops

frequecy [%]

Dose in Gy

-4

Study Group

Comparison between experimental and simulated fragment distributions after carbon irradiation.

0.006

-6

Human 3D Genome

0.004

0.002

-6

0.000 10.00

100.00

1000.00

size of DNA fragments [MBp]

10000.00

People

Human 3D Genome Study Group

Tobias A. Knoch

Tobias A. Knoch Carsten Mehring Christian Münkel Jörg Langowski

Biophysics of Macromolecules, German Cancer Research Center, Heidelberg, Germany

Steffanie Groß Karin Bütig Bernhard Horsthemke

Institute for Human Genetics, University of Essen, Germany

Irina Solovei Thomas Cremer

Institute for Anthropology and Human Genetics, University of Munich, Germany

Joachim Rauch Harald Bornfleth Christoph Cremer

Institute for Applied Physics, University of Heidelberg, Germany Karin Monier Kevin Sullivan

The Scripps Institute, La Jolla, USA

Angel Alonso

Applied Tumorvirology, German Cancer Research Center, Germany

Peter Lichter

Organisation of Complex Genomes, German Cancer Research Center, Germany

IBM-SP2, German Cancer Research Centre, Heidelberg Cray T3E, High-Performance Computing Center, Stuttgart IBM-SP2, Computing Centre, Karlsruhe Silicon Graphics-Graphic-Lab, Institute for Scientific Computing (IWR), Heidelberg The work is part of the Heidelberg 3D Human Genome Study Group which is part of the German Human Genome Project. We would like to thank the German Ministry for Sience and Technology (BMFT) for financing this project.

Approaches leading to the Three-Dimensional Organization of the Human Interphase Nucleus: Simulations, FISH, Chromatin Labelling in vivo, Fractal Analysis, Carbon Ion Irradiation

Knoch, T. A. LabFair 2000, Merck-Eurolab, Technology Centre Jülich, Jülich, Germany, 18th - 19th October 2000.

 

Abstract

Despite the successful linear sequencing of the human genome its three-dimensional structure is widely unknown, although it is important for gene regulation and replication. For a long time the interphase nucleus has been viewed as a 'spaghetti soup' of DNA without much internal structure, except during cell division. Only recently has it become apparent that chromosomes occupy distinct 'territories' also in interphase. Two models for the detailed folding of the 30 nm chromatin fiber within these territories are under debate: In the RandomWalk/Giant-Loop-model big loops of 3 to 5 Mbp are attached to a non-DNA backbone. In the Multi-LoopSubcompartment (MLS) model loops of around 120 kbp are forming rosettes, which are also interconnected by the chromatin fiber. Here we show with a comparison between simulations and experiments an interdisciplinary approach leading to a determination of the three-dimensional organization of the human genome: For the predictions of experiments various models of human interphase chromosomes and the whole cell nucleus were simulated with Monte Carlo and Brownian Dynamics methods. Only the MLS-model leads to the formation of non-overlapping chromosome territories and distinct functional and dynamic subcompartments in agreement with experiments. Fluorescernce in situ hybridization is used for the specific marking of chromosome arms and pairs of small chromosomal DNA regions. The labeling is visualized with confocal laser scanning microscopy followed by image reconstruction procedures. Chromosome arms show only small overlap and globular substructures as predicted by the MLS-model. The spatial distances between pairs of genomic markers as function of their genomic separation result in a MLS-model with loop and linker sizes around 126 kbp. With the development of GFP-fusion-proteins it is possible to study the chromatin distribution and dynamics resulting from cell cycle, treatment by chemicals or radiation in vivo. The chromatin distributions are similar to those found in the simulation of whole cell nuclei of the MLS-model. Fractal analysis is especially suited to quantify the unordered and non-euklidean chromatin distribution of the nucleus. The dynamic behaveour of the chromatin structure and the diffusion of particles in the nucleus are also closely connected to the fractal dimension. Fractal analysis of the simulations reveal the multi-fractality of chromosomes. First fractal analysis of chromatin distributions in vivo result in significant differences for different morphologies and might favour a MLS-modellike chromatin distribution. Simulations of fragment distributions based on double strand breakage after carbonion irradiation differ in different models. Here again a comparison with experiments favours a MLS-model.

Corresponding author email contact: [email protected]

Keywords: Genome, genomics, genome organization, genome architecture, structural sequencing, architectural sequencing, systems genomics, coevolution, holistic genetics, genome mechanics, genome function, genetics, gene regulation, replication, transcription, repair, homologous recombination, simultaneous co-transfection, cell division, mitosis, metaphase, interphase, cell nucleus, nuclear structure, nuclear organization, chromatin density distribution, nuclear morphology, chromosome territories, subchromosomal domains, chromatin loop aggregates, chromatin rosettes, chromatin loops, chromatin fibre, chromatin density, persistence length, spatial distance measurement, histones, H1.0, H2A, H2B, H3, H4, mH2A1.2, DNA sequence, complete sequenced genomes, molecular transport, obstructed diffusion, anomalous diffusion, percolation, long-range correlations, fractal analysis, scaling analysis, exact yard-stick dimension, box-counting dimension, lacunarity dimension, local nuclear dimension, nuclear diffuseness, parallel super computing, grid computing, volunteer computing, Brownian Dynamics, Monte Carlo, fluorescence in situ hybridization, confocal laser scanning microscopy, fluorescence correlation spectroscopy, super resolution microscopy, spatial precision distance microscopy, autofluorescent proteins, CFP, GFP, YFP, DsRed, fusionprotein, in vivo labelling.

Literature References

Knoch, T. A. Dreidimensionale Organisation von Chromosomen-Domänen in Simulation und Experiment. (Three-dimensional organization of chromosome domains in simulation and experiment.) Diploma Thesis, Faculty for Physics and Astronomy, Ruperto-Carola University, Heidelberg, Germany, 1998, and TAK Press, Tobias A. Knoch, Mannheim, Germany, ISBN 3-00-010685-5 and ISBN 978-3-00-010685-9 (soft cover, 2rd ed.), ISBN 3-00-035857-9 and ISBN 978-3-00-0358857-0 (hard cover, 2rd ed.), ISBN 3-00035858-7, and ISBN 978-3-00-035858-6 (DVD, 2rd ed.), 1998. Knoch, T. A., Münkel, C. & Langowski, J. Three-dimensional organization of chromosome territories and the human cell nucleus - about the structure of a self replicating nano fabrication site. Foresight Institute Article Archive, Foresight Institute, Palo Alto, CA, USA, http://www. foresight.org, 1- 6, 1998. Knoch, T. A., Münkel, C. & Langowski, J. Three-Dimensional Organization of Chromosome Territories and the Human Interphase Nucleus. High Performance Scientific Supercomputing, editor Wilfried Juling, Scientific Supercomputing Center (SSC) Karlsruhe, University of Karlsruhe (TH), 27- 29, 1999. Knoch, T. A., Münkel, C. & Langowski, J. Three-dimensional organization of chromosome territories in the human interphase nucleus. High Performance Computing in Science and Engineering 1999, editors Krause, E. & Jäger, W., High-Performance Computing Center (HLRS) Stuttgart, University of Stuttgart, Springer Berlin-Heidelberg-New York, ISBN 3-540-66504-8, 229-238, 2000. Bestvater, F., Knoch, T. A., Langowski, J. & Spiess, E. GFP-Walking: Artificial construct conversions caused by simultaneous cotransfection. BioTechniques 32(4), 844-854, 2002.