ABSTRACT
DNA sequences attaching loops of nuclear and mitochondrial DNA to underlying
structures in HeLa cells have been cloned and 106 representative clones
sequenced; 10 clones containing random genomic fragments served as controls. As
chromatin is prone to rearrangement, care was taken to isolate sequences using
`physiological' conditions that did not create additional attachments.
Comparison (by Southern blotting) of the concentration of each cloned sequence
in `total' and `attached' fractions of DNA showed that most clones did contain
attached sequences, but even highly-attached sequences were not attached in all cells in the population.
Results demonstrated that 28% of clones were derived from three specific parts
of the mitochondrial genome and 22% from different parts of the
alu
repeat. In addition, 41% of clones contained unique nuclear sequences; these
contained no more of the motifs found attached to nuclear scaffolds or matrices
(ie SARs or MARs) than would be expected from their base composition. No other
attachment motif(s) could be identified by sequence analysis. However, Northern
blotting showed that all the mitochondrial clones and 76% of clones containing
unique sequences were transcribed; the degree of attachment correlated with
transcriptional activity. These results are consistent with transcription being
responsible for ever-changing attachments in both nuclei and mitochondria.
It is widely assumed that nuclear DNA is looped into domains by attachment to
some underlying skeleton (e.g. a matrix or scaffold; for reviews see
1
-
3
). Most models for loop structure involve stable `structural' attachments of
repeated DNA motifs in most cells in an organism, as well as `functional'
attachments that vary from cell to cell, depending upon replicational and
transcriptional activity (
4
).
Three general approaches have been used to define attached sequences. One
involves progressively detaching DNA from the skeleton with a nuclease;
sequences at attachment points resist detachment and become enriched in a
pelleted fraction (
5
,
6
). Sequences with a high affinity for isolated skeletons can also be selected by
incubating isolates with various DNA fragments to see which bind specifically (
7
). And if attached sequences determine the functional boundaries of a chromatin
domain, they can also be defined by their ability to buffer the domain from the
repressive effects of flanking chromatin (
8
-
10
).
Despite the availability of these assays, there remains little agreement on the identity of the molecules at attachment points. Sometimes the
DNA sequences that resist detachment are found within genes (e.g. the canonical
matrix attached region, or MAR, of the kappa immunoglobulin gene;
7
), sometimes not (e.g. the canonical scaffold attachment region, or SAR, next to
the histone gene;
6
); sometimes MARs/SARs act as domain boundaries (e.g. the chicken lysozyme MAR;
9
), sometimes not (e.g. the MAR/SAR of the heat-shock locus;
10
). Moreover, different proteins with a high affinity for MARs/SARs have been
identified, including topoisomerase II (
11
), SATB1 (
12
), SAF-A (
13
) and ARBP (
14
). In the absence of any consensus, sceptics suggest that these differences
result from the induction by the unphysiological conditions used during isolation of new and artifactual attachments of loops to an underlying substructure (
2
,
15
).
Unphysiological conditions are often used because nuclei and chromatin tend to
aggregate at an isotonic salt concentration. However, aggregation can be prevented by encapsulating cells in agarose
microbeads (diameter 50-150 [mu]m) before lysis with Triton X-100 in a `physiological' buffer (
16
); then the now-encapsulated chromatin is accessible to molecular probes like enzymes and antibodies. The loops are sufficiently protected to retain their
attachments, integrity and contour length during extensive manipulation (
17
). We have now cloned and sequenced DNA fragments at attachment points in this
material. We find no evidence for `structural' attachments involving the same
(repeated) motif; rather, different sequences had different probabilities of
attachment, a high probability correlating with a high transcriptional
activity. Therefore attachments are predominantly `functional'.
Unsynchronized HeLa cells were grown in suspension for 24 h in [
3
H]thymidine (0.1 [mu]Ci/ml; ~50 Ci/mmol) to label uniformly their DNA; this allows the percent
chromatin remaining attached during isolation to be estimated from the percent
acid-insoluble [
3
H] remaining. Labelled cells were encapsulated in agarose (2.5 * 10
6
cells/ml packed beads), washed in a `physiological' buffer (PB; final
concentrations are 22 mM Na
+
, 130 mM K
+
, 1 mM Mg
2+
, <0.3 [mu]M free Ca
2+
, 132 mM Cl
-
, 11 mM phosphate, 1 mM ATP and 1 mM dithiothreitol; pH 7.4), permeabilized by
washing 3* in PB + 0.25% Triton X-100 and then rewashed 5* in PB. Beads, 2 ml (containing 5 * 10
6
cells), were resuspended in 10 ml PB, incubated (20 min; 33oC) with
Hae
III (5000 U) and
Eco
RI (25 000 U), washed in PB, reincubated with the two enzymes as before, and
released chromatin fragments electroeluted (3 V/cm; 5 h). Beads, 2 ml (now
containing 4.3% chromatin), were incubated in 10 ml PB with
Alu
I,
Hin
fI,
Rsa
I,
Pst
I (500 U each),
Hpa
I,
Msp
I,
Mbo
I and
Sau
3A1 (250 U each), placed in dialysis tubing, re-subjected to electrophoresis (2 V/cm; 4 h) and both beads and surrounding
fluid recovered. DNA was purified by treating beads with 0.2% sarkosyl, RNase A
(50 [mu]g/ml; 33oC; 30 min) and proteinase K (200 [mu]g/ml; 33oC; 1 h); after melting beads (70oC; 15 min), the solution was extracted successively with
phenol (this removes agarose), phenol-chloroform and then chloroform and DNA precipitated with ethanol. The
ends of fragments (number-average size 1.2 kb) were `in-filled' using `Klenow', ligated into Bluescript II vector (cut with
Sma
I and phosphatase-treated) and used to transform
Escherichia coli
strain DH5 to give ~5000 colonies (
18
; inserts had a number-average size of 1.0 kb). A similar library was also prepared from `total' DNA
(Fig.
1
, stage D).
For Figure
1
I, DNA was purified from (i) beads collected immediately after the second incubation with
Hae
III and
Eco
RI (total DNA,
Hae
III-
Eco
RI fragments; lane 1); (ii) beads immediately after the first electroelution (4.3% of
Hae
III-
Eco
RI fragments remaining attached; lane 2); (iii) buffer outside beads in the
dialysis bag after the second electroelution (2.9% released fragments; lane 3);
and (iv) beads in the dialysis bag (1.4% attached fragments; lane 4). The
number-average molecular weights of fragments and corresponding loop sizes were
determined (
17
,
19
) using agarose gels (1.7% in TEA buffer; 30 V per 10 cm, 5 h).
Sequences were determined after amplifying inserts in clones using the
polymerase chain reaction and Bluescript primers, using `Sequenase' (USB) and
6% polyacrylamide gels (
18
). Clones, 140, containing `attached' fragments (designated by C#) were
amplified; 106, 27 and 5 had inserts of <500, 500-2000 and >2000 bp, respectively (2 had no inserts; number-average size of inserts was 0.5 kb). The 106 with short inserts
were sequenced from both ends; most inserts were short enough to allow complete
overlap of the two resulting sequences (designated as LAS#), but longer ones
gave two partially or completely non-overlapping sequences (designated as LAS#A or B) that were analysed
independently. Sixteen inserts (8, 3 and 5 homologous to mitochondrial,
alu
and other sequences, respectively) were identical to others in the 106 clones
and so could have arisen by division of bacteria following the same initial
ligation event; they were not analysed subsequently. The resulting loop
attachment sequences (LASs) were analysed using the GCG package (program manual
for the Wisconsin GCG package, version 8) for homology with sequences in the
EMBL database version 44.25 were identical (or differed at 1-2 places) with mitochondrial sequences. Twenty, 3 and 3 clones were
homologous with
alu
, alphoid and LINE repeats, respectively. The TIGR database of ESTs (
20
) was then screened with the non-mitochondrial and non-repeated sequences, including unique sequences 5' or 3' to regions of homology with
alu
-a (named LAS#END1 and END2, respectively). Sequences were also analysed
using the GCG and Staden (
21
,
22
) suites, as well as PROMOTER SCAN (
23
), and deposited in the EMBL database with accession numbers X89762-7, X89831-40, X89842-8, X91562, X91844 (mitochondrial), X91537-61, X91563-85, X91587-610, X91841-4 (non-mitochondrial).
The relative concentrations of each LAS in `total' and `attached' DNA fractions
were determined by `Southern' blotting. Attached DNA was isolated as
illustrated in Figure
1
E after treating (20 min) beads (5 * 10
6
cells/ml beads) with
Hae
III (250 U/ml) and
Eco
RI (2500 U/ml). This modification was necessary as exhaustive cutting as in
Figure
1
E is expensive; as a result, ~90% chromatin is detached. Attached DNA was then purified as above, as was
total DNA after adding Sarkosyl to unencapsulated cells. `Total' and `attached'
fragments were completely recut with
Hae
III, precipitated, appropriate amounts of
3
H loaded on a gel, subjected to gel electrophoresis (all as above), blotted, hybridized,
autoradiograms prepared using X-ray film or a PhosphorImager, and enrichments of `attached' bands relative to those in `total' DNA
determined by quantitative densitometry (
24
).
32
P-Labelled probes (specific activity ~10
9
c.p.m./[mu]g) were prepared using a `random primed DNA labelling kit' (Boehringer).
Note that depletions cannot be quantitated accurately, as relevant
concentrations of `total' DNA were not run in adjacent tracks. Therefore all depletions (and enrichments of <2*) are (conservatively) grouped together; this has the effect of
underemphasizing the number of LASs that are attached.
The relative concentrations of sequences in the `total' clones was determined in
a way that overemphasised their attachment. Many of the
Hae
III fragments contained in the `total' library are <300 bp and so are lost during blotting. Therefore larger `total' and `attached'
fragments were prepared as in Figure
1
D and E using only
Eco
RI; ~20% chromatin was then retained and this was used as the `attached'
fraction for blotting. Then, a greater proportion of the sequences in the
genome appear enriched.
The concentration of transcripts in whole cells, nuclei and polyA
+
RNA that were complementary to LASs was determined using `Northern' blots (
18
). Cells were washed in PBS, and nuclei isolated by swelling cells (15 min, 4oC) in 50 mM NaCl, 1 mM MgCl
2
, 10 mM Tris (pH 8.0), breaking them with 20 strokes of a Dounce homogenizer,
before spinning (2000
g
, 5 min) and rewashing. RNA was extracted from cell or nuclear pellets using a
RNAzol B kit (Biogenesis); polyA
+
RNA was selected from whole cell RNA passing twice over oligodT-cellulose (Pharmacia). RNA samples were precipitated, redissolved in
formaldehyde gel-loading buffer and run (2 V/cm; 18 h) in 2% agarose-formaldehyde gels. Gels were either washed, stained with ethidium
and photographed (e.g. Fig.
5
, left) or RNA was transferred to nitrocellulose (0.45 [mu]m pores; Schleicher and Schuell) and hybridized with probes made as above;
finally autoradiographs were prepared. For Figure
6
, the strongest signal-whether in whole cell, nuclear, or polyA
+
RNA-was expressed on the scale reflecting the minimum exposure required to
detect a band on film (PhosphorImager exposures used for weak signals were
converted to an equivalent time for film): - (no signal), + (>5 d), ++ (16 h-3 d) and +++ (2 h).
Figure
1
illustrates the approach used to create a library containing sequences at
points of attachment. HeLa cells were encapsulated in agarose, lysed, and chromatin loops cut exhaustively with 10
different restriction enzymes; then most chromatin was electroeluted from the
agarose beads to leave only 1.4% of the original amount. Finally, DNA was
purified from these residual fragments and cloned. As mitochondrial remnants
remain associated with the cytoskeleton and resist elution (
25
), sequences involved in attaching the mitochondrial genome to the substructure
are also isolated.
This approach requires that existing attachments are not broken, nor new
attachments created, during the procedure; then, the contour length of the
loops should remain unchanged. Therefore we monitored the length of nuclear
loops, which can be calculated from the percentage of chromatin (i.e. [
3
H]DNA) remaining in beads and the length of attached fragments (
17
,
26
).
Eco
RI and
Hae
III cut cellular DNA in encapsulated cells into fragments that are multiples of 200 bp (Fig.
1
I, lane 1). These fragments have a weight-average molecular weight of 3.1 kb, the length expected if the enzymes cut
only in linker DNA (
19
). After removing most chromatin, the residual fragments are longer and the
nucleosomal repeat is less obvious (Fig.
1
I, lane 2); this is consistent with ~1 kb at attachment points being protected from cutting (
19
). A number-average molecular weight of 3.84 kb can be derived from the distribution
of fragments in lane 2 (see Materials and Methods) and-as 4.3% chromatin remained attached-the average contour length is then (100/4.3) * 3.84 = 89.3 kb. After trimming with eight other
restriction enzymes, released fragments are shorter, but retain an obvious
nucleosomal repeat of 182 bp (Fig.
1
I, lane 3). The attached fragments are also shorter, but are still ~1 kb longer than the released ones, with a less obvious repeat (Fig.
1
I, lane 4). The contour length also remains essentially unchanged, as the
reduction in length is offset by increased detachment.
These results show that throughout the procedure the contour length is close to
the average value of 86 kb obtained previously (
17
,
19
); therefore few attachments are made or broken during isolation. Note that
nucleosomes cannot `slide' along DNA because restriction sites that were
initially covered remain covered-and so uncut-during the lengthy incubations. Note also that as attached
fragments lack the obvious nucleosomal repeat typical of detached fragments,
they must be relatively free of contamination by them.
Clones, 140, were then selected and inserts in 106 sequenced from each end. (See
Materials and Methods for a discussion of the selection criteria.) Sixteen were
eliminated as they were identical to others among the 106; they probably arose
from repeated ligation of identical sequences in the original isolate, or as
bacteria divided during amplification of the library. The resulting LASs were
analysed for homology with sequences in the databases (Table
1
). Some proved to be mitochondrial,
alu
, alphoid or LINE repeats. Only two of the remainder were identified: LAS77
(part of the gene for the low-affinity Fc receptor; 27) and LAS95 (the non-transcribed region 5' to the rDNA locus; 28). The latter region has been shown to
be attached previously using this approach (
24
). Ten control inserts were also derived by fragmenting `total' DNA with
Hae
III and then cloning the fragments.
Table 1
Mitochondrial clones accounted for 28% (Table
1
), even though only ~0.15% DNA in a HeLa cell is mitochondrial (
29
). Clones tended to be derived from one of three regions of the genome (Fig.
2
). This probably results, in part, from an appropriate distribution of
Alu
I and
Hae
III restriction sites within the three regions. These two enzymes cut
efficiently under our conditions to generate 54 and 36%, respectively, of all
ends cloned by our procedure (not shown) and, as the resulting mitochondrial
fragments have the appropriate size, they are cloned preferentially and so over-represented in the library. But despite this efficient cutting, the three
regions were nevertheless retained.
Although
alu
repeats represent ~2.5-5% of the nuclear genome, 31% non-mitochondrial clones contained
alu
repeats (Table
1
). Despite this preferential attachment, no one region of the
alu
consensus sequence was invariably present (Fig.
3
). Most repeats were of sub-family a (
30
,
31
) but members of sub-families b (e.g. LAS05, 119) and perhaps c (LAS33) were also present; none
were `precise variants' (
32
). LAS23 was almost identical to a repeated unit downstream of an
alu
repeat consisting of eight tandem repeats each with an
Alu
I site at the same location (
33
). Three inserts homologous with alphoid DNA shared positions 76-121 in the consensus sequence (Table
1
;
34
), perhaps suggesting a common attachment point, but this region did not contain
the centromeric CENP-B box (positions 127-143;
35
).
We next confirmed that the library did indeed contain attached sequences.
Sequences remote from attachment points should be detached and so eluted from
beads, unlike those close to attachment points (
5
,
36
). Therefore we determined the relative concentrations of each cloned sequence
in `total' and `attached' DNA fractions by `Southern' blotting. As exhaustive
cutting with 10 enzymes is so expensive, the large quantities of `attached'
fraction needed were prepared using lower concentrations of
Eco
RI and
Hae
III; as a result, it contained only ~10% of the total. Equal weights of the two fractions were completely cut
with
Hae
III, resolved into discrete fragments by electrophoresis (e.g. Fig.
4
top, lanes 3,4), blotted on to a filter and probed with [
32
P]DNA from each clone in turn. Sequences close to attachment points will be
enriched in the `attached' fraction and so will yield bands of greater
intensity in the resulting autoradiograms (Fig.
4
bottom). The degree of enrichment is determined by reference to known amounts
of total DNA run in adjacent channels in the gel (Fig.
4
top, lanes 1-3).
The concentration of transcripts that were complementary to different LASs was
determined by `Northern' blotting. RNA from whole cells, nuclei or the polyA
+
fraction was separated by electrophoresis (Fig.
5
left, lanes 1-3, respectively), blotted and hybridized with
32
P-labelled probes prepared from clones containing either LASs or `total'
DNA; autoradiograms involving the same examples used in Figure
4
are illustrated in Figure
5
. Two repeats (i.e. LAS14 and 40) were not transcribed, but another (LAS81) gave
a strong signal with polyA
+
RNA and was the only LAS to hybridize with nuclear RNA. Of the single-copy sequences, LAS15 gave two weak bands in the polyA
+
fraction (arrows), but LAS73 was typical of most, giving a band in whole cell
RNA and a stronger band in the polyA
+
fraction. LAS77 (the Fc receptor) hybridized with two faint bands in the polyA
+
fraction (arrows). [cDNAs encoding the Fc receptor have been isolated from many
other libraries made from various tissues, showing that it is widely expressed
(
20
).] LAS117 gave strong bands and 124 was typical of several LASs (including 35,
93, 101, 124 and 130) that hybridized with bands of different sizes in whole
cell and polyA
+
RNA. The mitochondrial LASs, 8 and 62, are highly transcribed and give strong
signals after short exposures; transcripts from the latter, but not the former,
are poly-adenylated (as expected). (The lack of signal given by LAS8 with the polyA
+
and nuclear fractions reflects their purity.) None of the `total' clones tested
(e.g. 22) gave significant signals, even after long exposures using a sensitive
PhosphorImager; `total' clone 4 gave the strongest, but this was hardly above
background.
The results of this analysis are summarized in Table
1
. Figure
6
illustrates the rough correlation between the enrichment of both mitochondrial
and nuclear LASs and the concentration of their transcripts; control fragments
(shown in boxes) tend not to be enriched or transcribed.
We next screened the nuclear LASs containing unique sequences for possible
attachment motifs. The same sequences `shuffled' at random provided suitable
controls with the same base content. No LAS contained an exact match with the
(loosely-defined) topoisomerase II consensus sequence (
37
) and only seven contained the same sequence with <= 2 mismatches. [These values can be compared with 0 (exact match) and 9 (two
mismatches) given by the `shuffled' set.] One (none of the `shuffled' set) had <= 2 mismatches with the topoisomerase I consensus sequence (
38
). Eighteen (19 `shuffled') had <= 2 mismatches with the SAR A box (
39
), whilst 15 (15 `shuffled') had <= 1 mismatch with the SAR T box; 12 (11 `shuffled') shared these loose matches
with both boxes. Similarly, LASs and their `shuffled' counterparts contained
the same (few) number of SATB1 binding motifs (
12
). These results show LASs contain no more of these AT-rich motifs than are expected by chance. LASs were also screened (see
Materials and Methods) for (i) deviations in base composition, hairpin loops,
repeats, Z DNA, and common `words'; (ii) promoters; and (iii) the binding sites
for the transcription factors Sp1, AP2 and OCT1 that are known to be present in
HeLa cells (
40
); no significant differences were found between the LASs and their `shuffled'
counterparts.
We have isolated and cloned DNA sequences that attach nuclear and mitochondrial
DNA in loops to underlying structures. HeLa cells were encapsulated in agarose
beads to protect them during subsequent manipulation and permeabilized with
Triton X-100 in a `physiological' buffer; then loops were cut with restriction
enzymes, most DNA removed by electrophoresis and the residual 1.4% that
remained attached was isolated and cloned, and 106 representative clones
sequenced (Fig.
1
). It is important to note that these LASs are defined operationally by the
procedure used; any attachments sensitive to Triton or the electric field would
not be seen.
Chromatin is notoriously prone to rearrangement, and our approach requires that
few existing attachments are broken nor new attachments created during the
procedure. Therefore, we used isotonic conditions throughout. We have
previously shown that the chromatin fibre in these permeabilized cells remains
intact and that nuclear RNA and DNA synthesis can continue at roughly the rates
found
in vivo
(
16
); if chromatin aggregated artifactually on lysis we would expect to lose
activity. We also monitored the length of the nucleosomal repeat as well as the
contour length of the nuclear loops; both remained unchanged during isolation.
Moreover, nucleosomes did not `slide' along DNA under these conditions because
restriction sites that were initially covered remain covered, and so uncut,
during lengthy incubations. And as the attached fragments lacked the obvious
nucleosomal repeat typical of detached fragments, they must have been
relatively free of contamination by them. We also confirmed that we had indeed
cloned attached sequences by measuring the relative concentrations of
complementary sequences in `attached' and `total' DNA fractions (Fig.
4
); 54% of the clones tested but none of the controls behaved in the expected
manner (i.e. homologous sequences were enriched in the `attached' fraction;
Table
1
).
Mitochondria are associated with cytoskeletal elements that resist elution from
beads (
25
) and so we expected to clone mitochondrial sequences. However, 28% of our
clones were derived from this organelle (Table
1
), even though only ~0.15% of the DNA in the cell is mitochondrial (
29
). Even more surprisingly, clones tended to be derived from one of three
functionally important regions (Fig.
2
). One region includes the origin of replication of the heavy strand, the heavy- and light-strand promoters and the highly-transcribed rRNA genes, the second the other origin, and the
third a region that is commonly mutated or deleted in the mitochondrial
myopathies (
41
).
As discussed in the Introduction, two extreme kinds of attachment can be
imagined, `structural' and `functional', and we expected to find some
combination of the two.
Alu
sequences, which comprise 2.5-5% of the genome, constituted 31% of the non-mitochondrial LASs (Table
1
) and so become candidates for members of the `structural' class. However,
Southern blotting showed that most
alu
repeats were not preferentially associated with the substructure (not shown);
therefore only a fraction, presumably represented by the ones cloned, can be
involved in attachment. Moreover, no one region within the repeat was
invariably present in the
alu
LASs (Fig.
3
), so it seems unlikely that they contain a `structural' motif.
A `structural' motif that was permanently bound in all cells in the population
should be enriched 10-fold when all but 10% chromatin was detached. However, none of our nuclear
or mitochondrial LASs were enriched to this extent (Fig.
6
). The enrichments, which are generally 1.5-5*, imply that most LASs were attached in only a fraction (i.e. 15-50%) of loops in the population; they only have a higher
probability than others of being attached. By definition, then, they are
`functional'.
Although our LASs are characterized by their diversity, they do share one common
feature: most are transcription units and so again can be classified as
`functional'. For example, 76% unique LASs were transcribed compared to ~5% human genome (
42
) and only one of nine controls (Table
1
). In addition, all regions of the mitochondrial genome, and so all
mitochondrial LASs, are also transcribed (
43
). [This was confirmed for only two (Fig.
6
, LAS8 and 62).] Moreover, it is also possible, but difficult to prove, that
most
alu
LASs are transcriptionally active, since it is now known that many
alu
repeats are transcribed; they occur within other transcription units and some
are even functional polymerase III genes with internal promoters and operative
retinoic acid response elements (e.g.
44
-
46
). Even if the attached
alu
s are not transcribed, they nevertheless carry the stigmata of a transcription
unit. Strikingly, then, most of our LASs-whether they be unique, mitochondrial or
alu
sequences-turn out to be parts of transcription units.
Replicating sequences, and perhaps origins, are closely associated with nuclear
matrices and cages (
1
,
2
,
50
). Although <1% loops in our unsynchronized cells can be replicating at any one time, it
remains formally possible that origins mediate many attachments. However, it is
difficult to test this in the absence of appropriate assays.
There is an extensive literature on three kinds of nuclear substructure isolated
using hypo- and hyper-tonic salt concentrations, scaffolds, matrices and cages. Each is
different from the others and each is associated with a different subset of
genomic DNA (for reviews, see
2
,
15
). Our attachment sequences are unlike those associated with scaffolds (
3
); they contain no more topoisomerase II sites or SAR consensus sequences than
would be expected from their base composition. SARs and MARs are also very AT
rich, but our non-mitochondrial LASs were less AT rich (i.e. 54%) than the whole genome
(i.e. 60%). MARs are sometimes transcribed (
47
), sometimes not (
48
), depending on the precise procedure used for isolation; most of our LASs are
transcribed (Table
1
). Unlike MARs (
12
,
49
), our LASs contain no more inverted repeats or SATB1-binding sequences than `shuffled' controls. This lack of similarity with
either SARs or MARs is unsurprising as loops decrease in length during the
preparation of scaffolds and matrices (
17
); this means that the new attachments that create the smaller loops will
inevitably obscure pre-existing ones.
The attachments seen here, however, are similar to those seen earlier in cages
prepared by lysing HeLa cells in Triton and 2 M NaCl without prior isolation of
nuclei in hypotonic buffers. The resulting nucleoids contained loops of naked,
supercoiled, DNA attached to the cage; again mitochondrial DNA was
preferentially attached, no one repeated sequence was responsible for attaching
the nuclear loops, and attached unique sequences tended to be transcribed or
were rich in promoter or enhancer sequences (
50
). This similarity was expected as loop length increases slightly on isolation,
so it is unlikely that new, obscuring, attachments were created (
17
).
The attachments of a (nuclear) `minichromosome' containing two transcription
units have also been analysed using isotonic conditions and the approach
illustrated in Figure
1
(
51
). Two populations of minichromosomes were found; one was transcriptionally
inactive and could be eluted, the other was active but resisted elution. The
resistant minichromosomes were attached, on average, at only one point, either
through a promoter or a transcription unit; this suggested that they were bound
either through transcription factors (at a promoter) or an engaged polymerase
(on a transcription unit). Cutting attached minichromosomes with
Hae
III enabled most resulting fragments to elute without loss of polymerizing
activity; then the residual fragments still served as templates for engaged
polymerases. This suggested that minichromosomal attachments changed from
moment to moment as the template engaged an attached polymerase.
The simplest explanation for the results obtained here, and those obtained
earlier with cages and minichromosomes, is as follows. We now know that active
RNA polymerases are concentrated in HeLa nuclei in ~2100 discrete structures (diameter ~70 nm) attached to a nucleoskeleton (
52
,
53
; F. J. Iborra, A. Pombo, D. A. Jackson and P. R. Cook, manuscript submitted).
Each of these `factories' is associated with many loops and transcription
units. Initiation would involve attachment of a promoter first to transcription
factors and then a polymerase in a factory, before the transcription unit slid
through the polymerization site and nascent RNA was extruded into the factory;
after the completion of the transcript, the unit would detach from the factory.
At any moment individual templates would be at different stages in this cycle;
no one point would always be attached and different points would have different
probabilities of attachment, depending on promoter strength and polymerase
transit time. Enhancers (and any non-transcribed
alu
repeats) would also have a high probability of binding to a factory, but would
do so without initiating transcription. All such attachments are dynamic in the
sense that they continually change from moment to moment (
54
,
55
).
A similar kind of model can be extended to mitochondria. The nucleolar and
mitochondrial transcription systems share a common ancestry (
56
) and as nucleoli also contain transcription `factories' located on the
nucleoskeleton (
57
), it seems likely that mitochondria will contain homologous structures attached
to the cytoskeleton. Then the transcriptionally-active fraction of mitochondrial genomes would also be attached through
promoters or transcription units to the underlying structure.
We thank Sandra Smith and Liz Cowe for their help, and the Cancer Research
Campaign for support.
No. independent clones (%)
No. attached (%)
a
No. transcribed (%)
b
LASs
Mitochondrial
25 (28)
3/3
c
(100)
3/3
c
(100)
Nuclear
Unique
37 (41)
13/24
d
(54)
22/29
d
(76)
Alu
20 (22)
nt
e
nt
e
Alphoid
3 (3)
nt
f
nt
f
LINES
3 (3)
nt
nt
Other repeats
2 (2)
nt
nt
Controls
10 (100)
0/9
d
(0)
1/9
g
(11)
REFERENCES
Return