ABSTRACT
The murine poly(C)-binding protein (mCBP) was previously shown to belong to the group of K-homology (KH) proteins by virtue of its homology to hnRNP-K. We have isolated cDNA-splice variants of mCBP which differ by two variable
regions of 93 bp and/or 39
+-
3 bp respectively. Both variable regions are located between the second and
third KH-domain of mCBP. The characterization of a partial genomic clone enabled us
to propose a model for the generation of the second variable region by the use
of a putative alternative splice signal. The mCBP mRNA is expressed
ubiquitously and the protein is found predominantly in the nucleus with the
exception of the nucleoli. We have identified five proteins which interact with
mCBP in the yeast two hybrid system: mouse y-box protein 1 (msy-1), y-box-binding protein, hnRNP-L, filamin and splicing factor 9G8. The
interaction between mCBP and splicing factor 9G8 was confirmed
in vivo
. These results suggest a function of mCBP in RNA metabolism.
We have previously described the purification and characterization of the single-stranded (ss) DNA binding protein mCBP (murine poly(C)-binding protein) which was initially identified by virtue of its binding to a recombination hotspot in murine
retrotransposons (
1
). mCBP was shown to bind preferentially to poly(dC) and, to a lesser extent, to
poly(rC)
in vitro
. Cloning and characterization of one mCBP cDNA revealed a similarity to the
hnRNP-K protein (
2
). Both mCBP and hnRNP-K contain three KH (K-Homology) domains which were first described for the hnRNP-K protein (
3
). Furthermore, the KH domains are similarly distributed in both proteins.
However, the homology is restricted to the KH motifs, there is only an overall
homology of 36% at amino acid level between mCBP and hnRNP-K (
2
).
Several mCBP homologues have been isolated by other investigators. A murine cDNA
variant differing by 31 amino acids was described as hnRNP-X (
4
). Human homologues have been isolated by several groups (
5
,
6
). One of them, [alpha]CP-2 ([alpha] complex protein 2), was identified as a component of a
multiprotein complex stabilizing the human [alpha]-globin-mRNA (
6
). All of these proteins show a strong poly(C)-binding activity.
Interestingly, hnRNP-K also binds to poly(dC) and poly(rC)
in vitro
. It was described as the major C-binding activity in HeLa cells (
7
). hnRNP-K was identified as a component of the hnRNP complexes which assemble on
nascent hnRNA concurrent with transcription and which are thought to be
involved in RNA processing and RNA transport (
8
). Immunohistochemical studies revealed that hnRNP proteins are predominantly
localized in the nucleus (
9
) although some of them were also detected in the cytoplasm where they are
involved in mRNA stabilization (
10
). All members of the hnRNP family which have been characterized so far bind
specifically to RNA as well as to ssDNA
in vitro
(
11
). Binding was shown to be mediated by a RNA recognition motif (RRM) found in
most hnRNP proteins (
12
).
Although hnRNP-K is found in hnRNP complexes, it is not a typical member of the hnRNP
family as it lacks the classical RNA recognition motif. Binding of the hnRNP-K protein to single-stranded nucleic acids is mediated by its three KH domains (
13
).
KH domains appear to be highly conserved in evolution. They are found in many
species from
E.coli
to mammals (
14
-
16
) including the human FMR1 (fragile X mental retardation) gene product which is
associated with the Fragile-X-Syndrome (
13
). A reported mutation in one of the KH domains of the FMR1-protein impairs its RNA binding activity and results in mental retardation
(
13
,
14
). Other examples of KH proteins include Nova, a neuronal protein which is
recognized by an autoantibody of breast and lung cancer patients who develop a
disorder of motor control (
17
) and gld-1, a tumor suppressor gene identified in
Caenorhabditis elegans
(
18
). A common feature of many KH proteins is their ability to bind RNA and ssDNA
in vitro
(
2
,
7
,
13
,
19
). Altogether, there is increasing evidence that proteins containing KH domains
are involved in important regulatory mechanisms which in some cases are
associated with human diseases.
hnRNP-K exists as a group of highly related isoforms varying by small insertions
of five and/or 20 amino acids, respectively (
20
). These variants are likely to be generated by alternative splicing of a single
primary transcript-a phenomenon which seems to be a common theme among hnRNP proteins (
21
,
22
). Surprisingly, only one of the hnRNP-K variants was shown to act as a transcriptional activator of the human c-
myc
gene by binding in a sequence specific manner to a regulatory element upstream
of the c-
myc
promotor P1 (
23
,
24
). Furthermore, hnRNP-K has been shown to interact directly with the SH3 domains of s-src, lyn, fyn and the vav protooncogene product and thus may be
involved in translating cell surface receptor signals into regulatory events on
the level of gene expression and RNA processing (
25
-
27
).
At present, not much is known about the function of mCBP
in vivo
. In the present study, we describe the isolation of mCBP binding proteins by
the yeast two hybrid system as a step to elucidate the role of mCBP
in vivo.
Furthermore we report the isolation and characterization of several novel mCBP
splice variants.
A partial mCBP cDNA was isolated by RACE (rapid amplification of cDNA ends) with
degenerated primers based on the mCBP N-terminal amino acid sequence as described (
2
). The amplified product was subcloned into pBluescript II KS +/- and was used to screen a cDNA library (mouse liver/[lambda] ZAP vector, Stratagene) to obtain full length clones.
Nucleotide sequences were determined by double-stranded sequencing using the T7 Sequencing kit (Pharmacia) according to
the manufacturers instructions.
Poly A
+
RNA was isolated out of different mouse tissues (129 Ola) with the QuickPrep
Micro mRNA purification kit (Pharmacia) according to the manufacturers
instructions. Total RNA was extracted as described (
28
).
Poly A
+
RNA (1.5 [mu]g) or 15 [mu]g total RNA were size fractionated on 1% agarose/formaldehyde gels,
transferred to nylon membranes (Biodyne B, 0.45 [mu]m, PALL) and baked at 80oC for 15 min. Prehybridization and hybridization was performed in
hybridization buffer (50% formamide, 5* SSPE, 5* Denhardt's solution, 0.3% SDS, 0.3 mg/ml herring sperm DNA) at 42oC. Membranes were hybridized overnight with [[alpha]-
32
P]dCTP labelled mCBP-KL cDNA (Rediprime, Amersham) and then washed with 0.2* SSC/0.1% SDS at 62oC until excess probe was removed. After autoradiography,
membranes were stripped by washing with boiling 0.1% aqueous SDS solution and
rehybridized with a [beta]-actin cDNA probe as described above.
Poly A
+
RNA was digested twice with 10 U RNase free DNase (Promega). An aliquot of 5 [mu]g was used for first strand cDNA synthesis with 1 [mu]g of a dT
18
primer using Superscript II reverse transcriptase (Gibco, BRL). PCR reactions
(25 [mu]M of each primer, 1.5 mM MgCl
2
, 0.2 mM dNTPs, 2 U
Taq
polymerase, Promega) were performed as follows: 5 min 94oC; 1 min 94oC, 1 min 60oC, 2 min 72oC, 35 cycles; 10 min 72oC. Primers used in PCR:
A
, 5'-CAACTCAACTGAGCGGCAA-3';
Ai
, 5'-CTATTGCTGGCATTCCGCAG-3';
B2
, 5'-TCCATGTGCCTCAACCCTGA-3';
B1
, 5'-TCAGGG TTGAGGCACATGGA-3';
C1
, 5'-GGAGAGCTGGATTCAATGCC-3';
D
, 5'-CCAGACATCTGACGGATCTC-3';
Di
, 5'-TGATTTTGGCGCCTTGACGC-3'.
A mouse genomic cosmid library (Supercos1/129J) was screened with the entire
mCBP-KL cDNA. Colonies were transferred to nitrocellulose filters (Protran BA
85, 0.45 [mu]m, Schleicher & Schuell). The DNA was denatured (5 min, 0.5 M NaOH/1.5 M NaCl),
neutralized (5 min, 0.5 M Tris-HCl pH 7.5/1.5 M NaCl) and crosslinked to the membranes by baking at 80oC for 2 h under vacuum. Prehybridization and hybridization were
performed in aqueous solution at 68oC (3* SSC, 50 mM NaH
2
PO
4
pH 6.5, 0.5% SDS, 7 mg/ml non-fat milk powder and 3* SSC, 20 mM NaH
2
PO
4,
0.5% SDS, 7 mg/ml non-fat milk powder, respectively). Positive colonies were analyzed by
Southern blotting with the same probe. Hybridizing fragments were subcloned
into pBluescript II KS +/-. The exon/intron structure as well as the pseudogene sequence were
determined by double-stranded sequencing using primers derived from the cDNA sequence.
Jurkat cells were washed twice in PBS and resuspended in lysis buffer containing
150 mM NaCl, 50 mM HEPES pH 7.4, 1% NP-40, 1 mM PMSF, 1 mM NaF and 0.1 TIU/ml aprotonin (Sigma). Cells were
disrupted by short vortexing and incubation on ice for 15 min followed by a
centrifugation at 11 000
g
for 5 min at 4oC. The supernatant was used for immunoprecipitation.
Cellular lysates containing 500 [mu]g of proteins were precleared with 20 [mu]l Protein A/G PLUS agarose (Santa Cruz) followed by incubation for 2 h
at 4oC with 20 [mu]g purified polyclonal mCBP antibody. The agarose beads were washed
three times with lysis buffer and resuspended in SDS sample buffer. After
denaturation at 100oC for 5 min, proteins were size fractionated by PAGE and analysed by
western blotting.
The mCBP polyclonal antiserum was generated from guinea pigs, which were
immunized with the mCBP-KL protein fused to six N-terminal histidines. Immunoprecipitations were performed with 20 [mu]g of purified mCBP antibody or 5 [mu]l of purified anti maltose binding protein serum (Biolabs), 3 [mu]g/ml of the monoclonal 9G8 antibody (
29
) were used for the western blots.
Proteins were transferred on nitrocellulose membranes (Schleicher & Schuell) by semidry electroblotting and the membrane was blocked for 30 min in
2% non-fat milk powder in PBS. The first antibody was added in the same solution
and incubated for 16 h at 4oC, followed by three washes with 0.1% Tween-20 in PBS. The membrane was then incubated with a 1:10 000 dilution
of the POD-coupled second antibody (0.8 mg/ml, Dianova) and finally washed four
times. The presence of protein was detected by enhanced chemoluminescense
reaction (Amersham).
RV-SMC cells (smooth muscle cells of rat vein) were cultivated on cover slips
as described (
30
). After 10 min fixation at -20oC in methanol and 1 min at -20oC in acetone, cells were air dried for 10 min at room
temperature. Immunostaining was performed with a 1:100 dilution of anti mCBP
serum in PBS for 15 min at room temperature followed by a 10 min washing in PBS
and incubation of a goat anti guinea pig antibody coupled to Texas red
(Dianova) diluted 1:200 for 15 min at room temperature. After a final washing
for 10 min, cells were desiccated in ethanol and embedded in Mowiol. In
parallel, cells were stained with Hoechst DNA fluorochrome no. 33258 according
to standard procedures.
pGBT9-KL contains the full length cDNA of mCBP-KL cloned into the yeast expression vector pGBT9 (
31
). The yeast strain HF7c which contains two GAL4 inducible reporters (His3 and
LacZ) was transformed with the bait plasmid pGBT9-KL and then retransformed with a 9.5-19.5 day mouse embryo cDNA library. This library contains cDNA
inserts ranging from 350-700 bp generated by random primed cDNA synthesis (
32
). cDNA fragments were inserted into a pVP16 vector that directs the synthesis
of fusion proteins containing the mouse embryo peptides and the VP16 activation
domain. The plasmid pLEXA-Lamin (
32
), coding for a fusion protein between LEXA and human LaminC (amino acids 66-230), was used for yeast control transformations.
HF7c cells (
33
) were transformed to tryptophan prototrophy with pGBT9-KL and afterwards retransformed with the mouse cDNA library by standard
procedures (
31
,
34
). Histidine prototroph clones were assayed for [beta]-galactosidase (lacZ) activity by a filter assay (
35
). The library plasmids of the lacZ positive clones were recovered by isolation
of total DNA from the colonies followed by electroporation into
E.coli
and subsequent restriction analysis of both plasmids. The cDNA inserts were
sequenced by using the primer: 5'-GGTACCGAGCTCAATTGCGG-3' positioned in the polylinker region upstream the 5'-end of the cDNA insert.
We previously reported the isolation of an mCBP cDNA (referred to as mCBP-KL in this paper) by RACE followed by screening of a mouse cDNA library
with the RACE product (
2
). In the present study we analyzed additional mCBP cDNAs and detected two small
variable regions by comparing the sequences of two full length clones (mCBP-KL and mCBP-E, Fig.
1
A). Unlike mCBP-KL, mCBP-E contains a 93 bp insertion, whereas further downstream, 39 bp are
missing (Fig.
1
B). Both regions are located between the second and third KH-domain of mCBP. A characterization of additional mCBP cDNAs as well as the
RACE product pRACE 1.2 at these precise regions gave a more complex pattern
than initially expected. Concerning the first variable region of +-93 bp, all clones were identical either to mCBP-KL or mCBP-E whereas the second variable region showed a more diverse
phenotype: the maximum size of this region turned out to be 42 bp (compare mCBP-H versus mCBP-KL, Fig.
1
A) which have to be further divided into 39 + 3 bp as there are variants lacking
either the first 39 bp or the last 3 bp (mCBP-E and pRACE1.2, respectively). Provided that the +-39 +- 3 bp phenotype occurs independently from the first
variable region and that there are no further undetected variable regions, a
total of eight different mCBP variants (five of which have been characterized)
would exist, which are most likely generated by alternative splicing of a
single primary transcript.
Apart from the genomic clone described above, we obtained two cosmids whose
inserts hybridized to probes from the 5' and the 3' coding region of mCBP-KL (data not shown). The hybridizing fragments were
subcloned and analyzed for mCBP coding sequences. They contained two different
intronless mCBP sequences, which are highly degenerated compared with the cDNA
sequence (data not shown). This indicates that they represent processed
pseudogenes which resemble reverse transcribed copies of mRNAs integrated into
the genome (
37
). One of them resembles a mCBP cDNA with the -93 bp and -39 + 3 bp phenotype concerning the first and second variable
region, respectively, which is consistent with the model presented in Figure
1
.
mCBP expression was investigated by Northern blotting and showed the appearance
of a 1.9 kb transcript in mouse testes, thymus, brain and liver (Fig.
3
). Expression of mCBP was also confirmed in mouse lymph nodes and colon as well
as in EL4 cells (data not shown). Highest expression levels upon comparison
with actin expression were found in testes and EL4 cells whereas thymus, lymph
nodes, brain and colon showed lower mCBP expression. Only weak hybridization
was detected in liver.
A polyclonal anti mCBP serum was used to detect the subcellular localization of
mCBP in different cell lines. Mouse 3T3 cells as well as rat RV-SMC cells were immunostained with the mCBP and preimmune serum combined
with a concurrent staining of nuclei with a DNA fluorochrome. In RV-SMC cells and 3T3 cells the mCBP protein is localized predominantly to the
nucleus with the exception of the nucleoli (Fig.
5
and data not shown). During mitosis, mCBP is found dispersed in the cytoplasm,
where it migrates back to the nucleus during telophase.
Apart from the detailed characterization of the nucleic acid binding properties
of mCBP (
2
) and its identification as a component of a multiprotein complex stabilizing
the human [alpha]-globin mRNA (
6
), nothing is known about the function of mCBP
in vivo
. As nucleic acid binding proteins often exert their function by the interaction
with other proteins, we employed the yeast two hybrid system, originally
described by Fields and Song (
38
), to identify proteins which interact with mCBP.
A full length cDNA of mCBP-KL fused to the GAL4 DNA binding domain was used as a `bait' named pGBT9-KL. Yeast strain HF7c was retransformed with a 9.5-10.5 day mouse embryo cDNA library (see Materials and
Methods). The expression of mCBP in day 2.5 mouse embryonic stem cells as well
as in day 14 mouse embryonic fibroblasts was confirmed by EMSA (data not
shown).
A total of 4
* 10
6
transformants were placed on selective medium and then screened as described in
the Materials and Methods. False positives were excluded by transfecting HF7c
cells either with the library plasmid alone or together with pGBT9, pGBT9-KL or pLexALamin (negative control). Fifteen out of 400 clones showed [beta]-galactosidase expression only when transformed with pGBT9-KL and the library plasmid.
Sequencing of the 5'-ends of the 15 cDNAs which were identified in the yeast two hybrid
system revealed eight different sequences, three of them being unknown. The
remaining five cDNA fragments correspond to the following already described
sequences (Table
1
): msy-1 (bp 251-600), murine y-box-binding protein (bp 170-520), the murine homologues of human hnRNP-L (bp 220-580 ), human splicing factor 9G8
(bp 120-500) and human actin binding protein 280 (also known as filamin, bp 2400-2800).
Interestingly, four of the five cDNA inserts code for peptides involved in the
binding of nucleic acids. Msy-1 and y-box-binding protein both contain an 80 amino acid cold shock
domain, which is responsible for the binding of DNA and RNA. The isolated cDNA
fragment of murine hnRNP-L spans the region of its human counterpart containing one of the two RNP-1 motifs responsible for the binding of RNA. An RNP-1 as well as an RNP-2 motif is also present in the murine homologue of the
human splicing factor 9G8.
Table 1
The interaction between mCBP and splicing factor 9G8 was confirmed
in vivo
by co-immunoprecipitation. The specificity of the mCBP antibody was demonstrated
by Western analysis (data not shown). This antibody was then used to
precipitate mCBP from total cellular lysates as described in Materials and
Methods. Western blot was performed with the 9G8 antibody (Fig
6
).
The 35 kDa 9G8 protein co-purified with mCBP when the anti mCBP antibody was used (lane 2), whereas it was not co-immunoprecipitated with a mock antibody (lane 3). Jurkat cell extract served as a positive control (lane 1). These results strongly
suggest that the interaction between mCBP and the splicing factor 9G8 also occurs
in vivo
.
mCBP belongs to the poly(C)-binding proteins in vertebrates and, by the existence of three KH domains,
shows similarity to hnRNP-K, a member of the group of hnRNP proteins. In this paper, we have
described mCBP splice variants which differ by two variable regions of 93 bp
and/or 39 +-
3 bp respectively, both of them being located between the second and third KH
domain.
Characterization of a partial genomic mCBP clone revealed that the alternatively
spliced 3 bp do not represent an individual mini-exon. According to the model we present in Figure
2
, the usage of an alternative splice site within exon n-2 leads to the absence
of its first three bases. However, they do not represent a codon. The exons are
not spliced in frame so that the usage of the alternative splice site 4a leads
to the absence of the codon GCA which codes for alanine (Fig.
2
C). Alternative splicing of +-1 amino acid has also been described for hnRNP-K (
20
) and FBP, another member of the KH domain protein family (
39
); however, in none of these cases the genomic situation has been characterized.
The existence of multiple variants generated by alternative splicing is a common
feature of many hnRNP proteins: four different variants have been described for
hnRNP-K differing by two miniexons of 15 and/or 60 bp, respectively (
20
). Other examples are hnRNP-B1 and hnRNP-A2, which are identical except for 12 amino acids at their N-terminus (
22
), hnRNP-C1 and C2 which differ by 13 amino acids (
22
) and PTB, also described as hnRNP-I (
40
), which contains two variable regions of 19 and/or 7 amino acids respectively (
21
). The initial purification of mCBP (
2
) by ssDNA affinity chromatography resulted in two separate bands, which are
likely to represent two isoforms of the protein. Interestingly, both isoforms lack the 93 bp variable region, as determined by protein sequencing. This could indicate that
mCBP isoforms, which contain this region, do not bind ssDNA as strongly as the
mCBP isoforms without that region. This would be reminiscent of the hnRNP-K protein whose different splice variants seem to have different ss
nucleic acid binding properties as only one of the four isoforms was found to
interact with a regulatory element of the c-
myc
promotor (
23
). The ability of a protein to interact with both RNA and DNA is not unusual: in
Xenopus laevis
, the transcription factor TFIIIA, which activates the 5S rRNA transcription,
also binds the 5S rRNA transcript to form a 7S particle (
41
). Further experiments will have to be performed to determine if the different
mCBP variants have different nucleic acid binding properties.
Why do so many variants exist which differ only by a few amino acids? One
possibility is an alteration of their nucleic acid binding properties. The
observation that only one of the four splice variants of hnRNP-K has been identified as a transactivator of human c-
myc
gene (
23
) supports this hypothesis. Another possibility would be the modification of
protein-protein interactions, as hnRNP proteins are components of large
multiprotein complexes (
8
).
The expression of different mCBP variants was examined by Northern blotting and
RT-PCR. The same banding pattern of amplified DNA products was observed in
all tissues tested, which is consistent with the Northern blot results.
Interestingly, with the exception of one primer combination (Fig.
4
B, lane 9), we obtained more amplified PCR products as expected. The possibility
that these bands correspond to processed pseudogenes can be excluded because
the negative controls showed no amplified DNA. For that reason, it seems
possible that mCBP variants, in addition to those already described in this
paper, do exist. A human mCBP homologue has been described as [alpha]CP2 (
6
) and hnRNP-E2 (
42
). This protein contains an additional variation of four amino acids compared
with the mCBP clone pRACE 1,2 (Fig.
1
A) but is otherwise identical. These four amino acids are located within the
region amplified by RT-PCR. A comparable variant could also exist in mouse. This will have to be
confirmed by the isolation and characterization of further mCBP cDNAs. Another
human variant ([alpha]CP1/hnRNP-E1) has been characterized, which is only 89% identical to mCBP-E (
6
,
43
). Highest diversity is found dispersed in the region between the second and
third KH domain, which again is the region amplified by RT-PCR (Fig.
4
A). This variant was probably not detected by us because of the stringent
screening conditions we employed for the isolation of mCBP cDNAs. However, it
is possible that this and other variants could have been amplified in the RT-PCR assay.
In order to get insight into a possible function of mCBP
in vivo
, we employed the yeast two hybrid system to identify mCBP-binding proteins. Interestingly, five of the isolated sequences encode
parts of proteins which have been shown to participate in RNA-binding or metabolism. Both the msy-1 protein as well as the y-box-binding protein belong to the family of y-box-proteins originally identified by virtue of
their binding to promoters containing reverse CCAAT boxes. They have been
further demonstrated to participate in the packaging of RNA (
44
). msy-1 was shown to bind and stabilize paternal mRNA in mouse spermatocytes as
a component of the 60-80S mRNP particles (
45
). hnRNP-L is found to be associated with nascent transcripts of amphibian
lampbrush chromosomes (
46
) and binds to a
cis
-acting element of a HSV-TK transcript (
47
). Filamin (
48
) belongs to the group of actin crosslinking proteins. Because
in situ
hybridization revealed a co-localization of poly A
+
RNA, the actin crosslinking proteins are hypothesized to be involved in the
transport and positioning of mRNP particles (
49
). The splicing factor 9G8 belongs to the family of SR splicing factors (
29
) which are associated with spliceosomes that assemble on hnRNA after
transcription (
50
).
The interaction of mCBP with the splicing factor 9G8 was verified
in vivo
by co-immunoprecipitation (Fig.
6
B). Further experiments will reveal if the interaction of mCBP with msy-1, y-box-binding protein, hnRNP-L and filamin also occur
in vivo
. Interestingly, all of them are components of large multiprotein complexes
involved in RNA metabolism. In this context, it is interesting to note that the
human mCBP homologues [alpha]CP1 and [alpha]CP2 have also been characterized as parts of an mRNA stability
complex which binds to the 3'-UTR of [alpha]-globin-mRNA in erythrocytes (
6
).
The above results favour a participation of mCBP in RNA metabolism whereas mCBP
was described as a ssDNA binding protein based on its preferential interaction
with poly(C) sequences
in vitro
by Goller
et al
. (
2
). This is not necessarily a contradiction as there are examples for
bifunctional proteins, including hnRNP-K, which bind to RNA as well as to DNA (
7
,
23
,
41
).
We thank Dr Michael Zimmer for providing the cosmid library and for helpful
discussion. We also thank Dr Andreas Kalmes and Carsten Hagemann for their
support with the Yeast Two Hybrid System. This work was supported by the
Deutsche Forschungsgemeinschaft (SFB 165).
mCBP binding
Characteristic protein
EMBL
proteins
domain
accession no.
msy-1
CSD
M62867
y-box-binding protein
CSD
L35599
hnRNP-L
RNP-CS
X16135
splicing factor 9G8
RNP-CS
L22253
filamin
-
X53416
REFERENCES
Return