ABSTRACT
We have purified to near homogeneity a novel nuclear protein from HeLa cells,
that specifically binds to scaffold or matrix attachment region DNA elements
(S/MAR DNA). The protein, designated SAF-B for scaffold attachment factor B, is an abundant component of chromatin, but not of the nuclear matrix and is expressed in all human
tissues investigated. Antibodies against the purified protein were raised in
rabbit and used to isolate the complete cDNA encoding SAF-B by immunoscreening. As predicted from the cDNA sequence, SAF-B contains 849 amino acids (96 696 Da), without significant homology
to any known protein. SAF-B is rich in charged residues, leading to an aberrant migration on SDS
gels, and has two putative bipartite nuclear localisation signals.
Eukaryotic chromatin is organised in a higher order structure consisting of
thousands of discrete, topologically constrained loop domains. These loops are
fixed at their bases to a network composed of proteins and RNA that is
generally referred to as the nuclear matrix or scaffold (
1
,
2
). Increasing evidence implies that a tight binding of chromatin to the nuclear
scaffold is not only important for the compaction of the chromatin fiber, but
is also involved in many aspects of nucleic acid metabolism (see
3
for review). It is widely accepted that loop domains are the units of gene
expression and replication, and are thus possibly also involved in the
formation of nuclear subcompartments (
4
-
6
). Attachment of chromatin to the nuclear scaffold seems to occur via
specialised DNA elements that have been found in all eukaryotic organisms
investigated. Termed SARs or MARs (for scaffold or matrix attachment regions;
in the current study we call them S/MARs), these DNA elements are
evolutionarily conserved, as shown for example by the fact that mouse S/MARs
bind to yeast nuclear scaffolds (
7
). Consequently, the DNA regions conferring chromatin attachment to the scaffold
are of considerable interest and have thoroughly been investigated in the last
decade. In many cases, S/MAR elements co-map with boundaries of actively transcribed chromatin domains, and are
postulated to protect the transcribed domain from regulatory mechanism from
neighboring sequences (
8
,
9
). However, the initial interpretation that S/MARs generally form the borders of
transcribed regions (and thus delimit units of gene expression) has been
questioned by the discovery of intronic S/MARs (e.g.
10
-
12
). These intronic S/MARs are indistinguishable from gene-flanking S/MARs with respect to their nucleotide sequence, their
interaction with scaffold preparations and their ability to confer position
independent transcriptional activation. It is therefore likely that both types
of S/MARs perform the same function
in vivo
, the anchorage of chromatin loops to the nuclear scaffold, and, presumably,
regulatory effects on adjacent genes.
Generally, S/MARs are DNA fragments of 300-3000 bp length that contain several A+T rich sequence motifs and
sequences resembling topoisomerase II cleavage sites (
13
,
14
). However, an interesting result from the comparison of the high number of
characterised S/MARs is the fact that no simple consensus sequence seems to
exist for nuclear scaffold attachment (
13
). Although several short A+T rich sequence motifs are clustered in most S/MAR
DNA elements, no single one of these sequences is characteristic for all
S/MARs. Rather, the binding of S/MARs to the nuclear scaffold is highly
dependent on both the A+T-richness and the length of the DNA fragment, indicating that an
interaction is involved that is strikingly different from well characterised
DNA-protein interactions as, for example, with transcription factors or
restriction enzymes. At present, however, the mechanism by which the nuclear
scaffold recognises the S/MAR DNA elements is not understood. It has been
proposed that the nuclear scaffold contains proteins that specifically
recognise unusual DNA structures such as tracts with a narrow minor groove (
15
), the single-stranded status of `unwinding elements' (
16
-
18
) or DNA bends (
19
). Further insight into the underlying recognition mechanisms, however, is only
possible by identifying and characterising these proteins in molecular detail.
A general assay to screen for proteins with a possible function in the anchorage
of chromatin loops is the use of a S/MAR DNA element as molecular affinity
probe in Southwestern blot procedures or in direct cDNA screening. Both approaches have been used
successfully to identify proteins of the desired specificity (
20
,
11
,
21
). Unfortunately, the few well characterised S/MAR-binding proteins, among those histone H1 (
22
), topoisomerase II (
14
), lamins (
23
,
24
), SATB1 (
21
) and SAF-A/hnRNP-U (
11
,
25
,
26
) show no homologies on the level of nucleotide or amino acid sequence. It is
therefore not yet possible to say what property of S/MAR DNA is specifically
recognised by these proteins. To gain insight into the molecular mechanism of
S/MAR DNA recognition, we have set out to identify and characterise proteins
that specifically interact with S/MAR DNA elements. In our search, we have
previously identified four nuclear proteins from HeLa cells with the desired
specificity for S/MAR DNA (
11
). These proteins were termed scaffold attachment factors A through D (SAF-A to SAF-D), dependent on their relative abundance in nuclear extracts. The protein SAF-A, characterised in our original publication (
11
), was later shown to be identical to the protein hnRNP-U (
25
,
27
-
29
), and the specific binding of this protein to S/MAR DNA was independently confirmed by others (
26
,
30
). In this communication we report on the purification, cloning and
characterisation of the second abundant protein SAF-B, the scaffold attachment factor B.
All buffers contained 10 mM mercaptoethanol to protect free thiol groups and 10
mM Na
2
S
2
O
5
(sodium metabisulfite, buffered to pH 8.0 with NaOH) as a general protease
inhibitor; all purification steps were carried out in the cold. For the
purification of SAF-B, nuclear extract was prepared from 1 * 10
10
HeLa S3 cells (obtained from Computer Cell Culture Center, Brussels) as
previously described (
11
). Briefly, cells were washed in phosphate buffered saline, allowed to swell for
10 min in hypotonic buffer (10 mM Tris-HCl pH 8.0, 10 mM NaCl, 3 mM MgCl
2)
and broken by 20 strokes in a loose fitting dounce homogeniser. Nuclei were
pelleted (750
g
, 10 min, 4oC), washed three times in the same buffer and then extracted into 80 ml of
extraction buffer (10 mM Tris-HCl pH 8.0, 500 mM NaCl) for 10 min. Nuclear debris is removed by
ultracentrifugation (150 000
g
, 30 min, 4oC) and the cleared extract is directly mixed with 20 ml of pre-equilibrated hydroxylapatite. After 30 min on a rocking platform, the
hydroxylapatite is pelleted (160
g
, 2 min, 4oC), washed twice in 50 ml of binding buffer and twice in 50 ml buffer
containing 70 mM potassium phosphate. Bound protein is eluted in 80 ml elution
buffer (170 mM potassium phosphate), diluted 2-fold with 10 mM Tris-HCl pH 8.0, 10 mM Na
2
S
2
O
5
, and directly loaded on a pre-equilibrated 10 ml Mono Q column (Pharmacia) at a flow rate of 1 ml/ml.
Bound protein is eluted in a linear gradient from 100 to 700 mM NaCl in
dilution buffer with a total volume of 60 ml. Fractions containing SAF-B are determined in a Southwestern blot assay with the labelled S/MAR
element MII and a 1000-fold excess of unlabelled
E.coli
DNA as unspecific competitor. SAF-B elutes at ~300 mM NaCl from Mono Q. Active fractions are pooled, diluted 4-fold and applied to a 1 ml Mono S column (Pharmacia) at a flow
rate of 0.5 ml/min. Elution is carried out in a linear gradient from 50 to 700
mM NaCl, where SAF-B elutes between 200 and 350 mM NaCl. For the final purification, the
eluate from the Mono S column, containing ~150 [mu]g SAF-B, is diluted 4-fold, mixed with 0.5 mg
E.coli
DNA, incubated on ice for 15 min and then pelleted (8000
g
, 15 min, 4oC). The pellet was resuspended in high salt buffer (10 mM Tris-HCl pH 8.0, 400 mM NaCl) and passed through a DEAE-Sepharose column to remove the DNA. The flowthrough,
containing ~100 [mu]g of nearly homogenous SAF-B, is mixed with the same volume of glycerol and stored at -20oC.
Nearly homogeneous SAF-B or the bacterially overexpressed partial protein p43 was further
purified by SDS-PAGE and subsequent copper chloride staining to remove remaining
impurities. The desired bands, containing 100-200 [mu]g of protein per immunisation, were excised, destained and
homogenised by 10 passages through a 25 gauge cannula. The homogenised material
was mixed with Freund's complete adjuvant for the first immunisation or RAS
adjuvant for the two boost immunisations after 6 weeks each. The obtained
antisera were made monospecific by affinity purification with the purified
antigen before further use according to (
31
).
DNA binding was assayed by a Southwestern blot procedure as previously described
(
11
) with DNA probes end-labelled radioactively by Klenow polymerase.
Polyclonal affinity purified serum against the SAF-B protein was used for immunoscreening of a HeLa LambdaZAP II library
(Stratagene #937216) as described (
32
). Several positive clones were isolated and sequenced by the dideoxy method (
33
) using the USB sequenase kit. A 540 bp probe from the 5' end of the longest clone was used for conventional hybridisation
screening of the same library. Among the 14 positive clones, one contained the
complete coding region along with the 5' and 3' non-translated regions. This clone, named A5, was sequenced from
both strands as described above.
The complete cDNA clone A5 was
in vitro
translated with the TNT rabbit reticulocyte lysate kit (Promega) in the
presence of [
35
S]methionine (Amersham) according to the manufacturer's suggested conditions.
A fragment of the complete cDNA encoding amino acids 443-718 of SAF-B was introduced in frame into the pRSET prokaryotic expression vector by conventional cloning procedures (
32
). Positive clones were identified by DNA minipreparations and assayed for
protein overexpression after induction with 1 mM IPTG for 2 h. Overexpressed
protein, named p43 according to its apparent molecular weight in SDS gels, was
purified by metal chelate chromatography over Ni-Agarose and used for the
production of antibodies.
SDS-PAGE of proteins was performed according to Laemmli (
34
); the gels were stained with silver (
35
), with Coomassie brilliant blue (
32
) or copper chloride (
36
). Western transfer was performed according to (
37
) with affinity purified polyclonal serum (
31
), alkaline phosphatase coupled secondary antibodies (Sigma) and BCIP/NBT
substrate. Protein concentrations were determined using the BioRad protein
assay reagent with bovine serum albumin as standard. DNA was labelled with [
32
P]dATP using Klenow polymerase according to standard protocols (
32
). Northern blotting was performed on blots obtained from Clonetech according to
the conditions suggested by the manufacturer. Cell fractionation experiments
were done as described previously (
25
).
The protein SAF-B was first identified by Romig
et al
. (
11
) as one of four proteins that specifically bind to the S/MAR DNA element MII
from the human topoisomerase I gene locus (
12
). Subsequently, we have developed a purification protocol for this protein. SAF-B was monitored throughout column chromatography by its ability to bind
radioactively labelled S/MAR DNA in the presence of a vast excess of unspecific
competitor DNA. This assay is highly specific for S/MAR-DNA binding proteins, although several abundant S/MAR binding proteins
such as topoisomerase II and HMG I/Y escape detection, most probably because of
their inability to refold to an active conformation after denaturing gel
electrophoresis. Our purification procedure starts from a 500 mM NaCl nuclear
extract prepared from 1 * 10
10
HeLa cells. The nuclear extract (Fig.
1
, fraction NE) was bound to hydroxylapatite material in a batch procedure. The
eluate at 170 mM potassium phosphate (Fig.
1
, fraction HAP) was diluted and directly applied to a FPLC Mono-Q column. Unbound proteins were washed off and the column was eluted with
a linear gradient from 100 to 700 mM NaCl. Active fractions with the peak of
SAF-B at ~300 mM NaCl were combined (Fig.
1
, fraction Mono-Q) and loaded to a FPLC Mono-S column. SAF-B eluted from Mono-S between 200 and 350 mM NaCl (Fig.
1
, fraction Mono-S). After chromatography on Mono-S, SAF-B is the main protein in the active fractions. However, it
still contains the activity of SAF-C, a protein with a molecular weight of 100 kDa. Removal of SAF-C and other contaminating proteins was achieved by exploiting the
ability of SAF-B to form aggregates in the presence of nucleic acids. After
centrifugation, the SAF-B/DNA aggregates were disrupted in high salt buffer and DNA was removed by
a passage over DEAE-Sepharose, resulting in nearly homogeneous SAF-B as judged by silver staining (Fig.
1
, fraction aggregation). On average, the purification yields ~100 [mu]g of SAF-B protein from 10
10
cells, with a recovery of 15-20% as estimated in Southwestern blots by comparing activity of SAF-B in total cell extracts of a known number of cells with a known
amount of purified protein. Purified SAF-B migrates as a single band of 150 kDa under denaturing conditions and is
a monomeric protein with an apparent sedimentation coefficient of 4.2S
20,w
in non-denaturing glycerol gradient centrifugation. Based on its apparent
molecular weight, we speculated that SAF-B could be identical or related to topoisomerase II, a protein with known
binding specificity for S/MAR DNA (
14
). However, purified SAF-B is free of topoisomerase activity and is not recognised by antibodies
against purified topoisomerase II in Western blotting experiments.
Additionally, the cDNA sequence of SAF-B, reported later in this paper, reveals no homology to topoisomerase
sequences.
As described above, SAF-B was purified from nuclear extracts by virtue of its specific binding to
the S/MAR DNA element MII in the presence of a 1000-fold excess of unspecific
E.coli
DNA as a competitor. To confirm the binding specificity of purified SAF-B, we repeated Southwestern blot experiments with the labelled S/MAR DNA
element MII (Fig.
2
A) and the non-S/MAR DNA pUC18 of similar length (Fig.
2
B). We find that SAF-B binds approximately equally well to both DNAs in the absence of
competitor DNA, indicating that SAF-B has a general DNA binding activity. However, with increasing amounts of
E.coli
competitor DNA, pUC18 is readily displaced from SAF-B, while ~40% S/MAR DNA remains bound to SAF-B even in the presence of a 2000-fold excess of competitor DNA (Fig.
2
C). With even higher amounts of competitor, binding is gradually lost, until no
binding is detectable at a 25 000-fold excess of
E.coli
DNA (data not shown). Specific binding was also observed with several other,
heterologous S/MAR elements like B1X1 or B4B5 from the chicken lysozyme gene
locus (
38
; a gift of Dr W. Strätling, Hamburg) or fragment IV from the upstream S/MAR of the human
interferon-[beta] gene locus (
39
; a gift of Dr J. Bode, Braunschweig) (data not shown). We conclude therefore,
that SAF-B is a DNA binding protein with high specificity for S/MAR DNA elements.
In experiments with restriction fragments derived from different S/MARs, we
find that SAF-B has no easily defined consensus binding site, but that specific binding
is dependent on both A+T richness and length of the subfragment (data not
shown). For the purpose of this paper, however, we do not particularly focus on
a detailed characterisation of the DNA binding properties of SAF-B.
To enable screening for the cDNA encoding SAF-B, we developed a polyclonal antiserum against highly purified SAF-B. The serum obtained recognised SAF-B in both crude nuclear extracts and in its purified form.
Specific antibodies were affinity purified by binding to immobilised SAF-B and were used for immunoscreening a Lambda ZAP expression library.
Several positive clones were isolated and sequenced. As none of the isolated clones contained the 5' end of the cDNA, a 540 bp probe from the 5' end of the longest clone was used to rescreen the same library by
DNA hybridisation. Fourteen positive clones were isolated, one of which was 2.8
kb in length and contained the complete coding region of SAF-B (sequence deposited to the EMBL, GenBank and DDBJ databases under
accession number L43631). To verify that this clone, termed A5, was a cDNA
clone encoding the SAF-B protein, we synthesised [
35
S]methionine-labelled protein in a coupled
in vitro
transcription-translation system. The
in vitro
synthesised protein showed the same electrophoretic mobility as authentic,
purified SAF-B from HeLa cells (Fig.
3
A). Additionally, antibodies produced to the bacterially overproduced partial
protein p43 (amino acids 443-718 of SAF-B) recognise SAF-B in unfractionated HeLa nuclear extracts and in the purified
form (Fig.
3
B). Taken together, these facts strongly indicate that the A5 cDNA clone encodes
the full length SAF-B protein.
We have focused on SAF-B as a DNA binding protein because of its specificity for S/MAR DNA, which makes it a good candidate for a
protein located at the attachment point of chromatin loops
in vivo
. Several criteria should be fulfilled by such a protein, and were addressed
experimentally. First, a protein with a structural function in the nucleus
should be abundant. From our purification that yields ~100 [mu]g of SAF-B from 10
10
HeLa cells with a recovery of 15-20%, we calculated a copy number of ~10
5
molecules of SAF-B per nucleus. Although this copy number is lower than that of other S/MAR
binding proteins (e.g. histone H1 or SAF-A/hnRNP-U), it is compatible with a structural function of SAF-B. Secondly, a general S/MAR binding protein should be
ubiquitous, i.e. expressed in all or the majority of cells and tissues. We have
performed a Northern blot analysis with a part of the SAF-B cDNA clone A5 as a probe (Fig.
5
). The left panel of Figure
5
B shows a Northern blot with poly(A)
+
RNA from a collection of neoplastic cells of different origins, the right panel
is a blot with poly(A)
+
RNA from healthy human tissues. A specific mRNA species with a length of 3.4 kb is detected in all cells and tissues at approximately the
same level, consistent with SAF-B being a ubiquitous (housekeeping) protein, at least on the level of gene
expression.
In higher eukaryotes, the genomic DNA of a few meters in length has to be
compacted in some way to be confined within the nucleus of only some
micrometers in diameter. This impressive compaction is brought about by the
formation of chromatin, whose basic architecture-DNA wound around octamers of histone proteins-is well understood. However, many details on higher order
structures of chromatin remain obscure. A new area in chromatin research was
initiated by the discovery of a proteinaceous nuclear framework (
46
) and subsequently by the identification of DNA fragments that specifically bind
to this framework (
47
,
48
). Combining data from electron microscopic examination of chromosomes (
49
,
50
) and biochemical work on chromatin structure (
51
,
47
), Gasser and Laemmli (
1
) have proposed a model according to which chromatin is organised in
constrained, topologically independent loops attached to a structural entity designated nuclear matrix or scaffold. Although this model is widely accepted today, little is known about the
mechanism of attachment of chromatin to nuclear substructures. To gain insight
into this functionally important aspect of nuclear structure, DNA elements
located at attachment points have been used in several laboratories to identify
nuclear proteins that could be involved in the formation of chromatin loop
domains (
20
,
11
,
23
,
21
,
26
).
In this communication, we describe the purification and cDNA cloning of a novel
protein, designated the scaffold attachment factor B (SAF-B). We have identified the SAF-B protein as one of four proteins in HeLa cell nuclear extracts that
specifically interact with S/MAR DNA, and consider it a candidate protein for a
molecular anchor at the basis of chromatin loops. Biochemical fractionation of
cells demonstrate that SAF-B is a chromatin protein, but not a constituent of nuclear matrix
preparations. This finding is interesting, as there is a general notion that a
protein involved in S/MAR DNA attachment should be part of the insoluble
substructure of nuclei. However, at least one other protein with significant
binding specificity to S/MAR, namely histone H1, is also not present in nuclear
matrix preparations but in negligible amounts (
22
) and is bound to chromatin even weaker than SAF-B. It is thus possible that two types of S/MAR DNA binding proteins exist,
which differ in their partitioning upon biochemical fractionation, but are both
involved in S/MAR element function
in vivo
.
As expected for a protein with a general function in chromatin structure, SAF-B is expressed equally in all cell types investigated, suggesting a
housekeeping nature of the protein. This behavior is reminiscent of SAF-A/hnRNP-U and histone H1, other known S/MAR binding proteins, but different
from SATB1 or topoisomerase II which are differentially expressed in different
cell types (
21
,
52
).
The complete cDNA for SAF-B, obtained by immunoscreening with antibodies developed against the chromatographically purified protein, revealed that SAF-B is a unique protein with no homology to known proteins. It
contains highly charged regions, with both the N- and the C-terminus being basic and the central half being acidic. SAF-B has two putative nuclear localisation signals, compatible
with our finding that the protein is located in the nucleus. Although SAF-B is clearly a DNA binding protein, computer comparisons to EMBL and
GenBank databases found no significant homologies to any previously identified
DNA binding protein. We can therefore not yet define by analogy which part of
the protein confers DNA binding. It could be argued that DNA binding of SAF-B occurs due to non-specific electrostatic interaction between basic regions of the
protein and the negatively charged phosphate backbone of DNA. However, a simple electrostatic interaction
would not be consistent with the specific binding of SAF-B to S/MAR DNA elements.
Although no homology to other cloned S/MAR binding proteins are evident on the
level of the amino acid sequence, SAF-B shares the ability for nucleic acid dependent self-aggregation with histone H1 (
22
), topoisomerase II (
14
) and SAF-A/hnRNP-U (
25
). At present, we do not know the molecular basis of the aggregation of these
proteins, but it is likely that both a DNA binding domain and a protein-protein interaction domain are involved in this process. It is thus
possible that S/MAR DNA binds to protein aggregates due to its intrinsic
flexibility brought about by interspersed A+T rich elements. This flexibility
could allow the DNA to follow the path of protein aggregates, and could
discriminate S/MAR DNA from bulk genomic DNA. Such a binding mode requires
strong protein-protein interactions, while the binding of a protein monomer to DNA could
be comparatively weak and unspecific. This model could be an explanation why
S/MARs have to be of certain length (usually >500 bp) to be specifically bound.
Future experiments will focus on the domain structure of SAF-B, with the aim to identify protein regions responsible for specific DNA
binding and protein-protein interaction.
After our manuscript was accepted, we learned that Drs. Jean-Pierre Bourquin and Walter Schaffner (Molecular Biology, University of
Zurich, Switzerland) had independently cloned the same cDNA. Sequence
comparison revealed a critical difference at position 344, resulting in a
frameshift concerning the first 100 amino acids of SAF-B. This error has been corrected in the database entry. Additionally, we
conclude that the 5' end of the SAF-B cDNA is missing in our clone A5. We apologize for the error and
any confusion it might have caused.
The authors wish to thank Dr Arndt Richter for encouragement and helpful
discussions at the beginning of this work, Annette Hecker for excellent
technical assistance and Dr Rolf Knippers for support and critical reading of
the manuscript. This work was supported by the Deutsche Forschungsgemeinschaft
through SFB 156.
REFERENCES
Return