The concept of the nuclear matrix, a karyoskeletal structure that serves as
a support for the genome and its activities, has stimulated many studies o
f the association of nuclear components and functions with this structure.
However, certain experimental findings are not consistent with the existenc
e of the nuclear matrix in vivo, including our inability to visualise a cor
responding structure in intact cells, the demonstrated mobility in vivo of
chromatin and messenger ribonucleoprotein particles, which are claimed to b
e bound to the nuclear matrix, the paradoxical extractability from nuclei i
n low ionic strength buffers of enzymes that are found in the 2 M NaCl-inso
luble matrix, and the extractability, in conditions which reproduce the int
ranuclear milieu, of regions of DNA (matrix or scaffold attachment regions,
MAR/SARs) postulated to be bound to the nuclear matrix in vivo. This revie
w considers the nuclear matrix model in the light of sometimes overlooked e
vidence that each step in its isolation may cause nuclear components to bin
d to it by new liaisons that do not exist in vivo. This is illustrated by e
xperiments where nuclear-targeted green fluorescent protein is found in the
nuclear matrix, and raises the possibility that MAR/SARs actually bind to
DNA-binding proteins or multiprotein complexes, including replicational, tr
anscriptional and processing machinery, and topoisomerases that are incorpo
rated into the nuclear matrix during its preparation. Considering that the
nuclear lamina forms a rigid exoskeleton, the necessity for internal skelet
al structures is raised; the major roles that macromolecular crowding, phas
e partitioning, and charge effects are likely to play in organisation of th
e intranuclear space may provide new models for the compartmentalisation of
proteins and functions into different nuclear domains and of chromosomes i
nto territories.