The p20 protein was first detected in HeLa cells as a protein interacting with the GCC-element of rpL32 [18], and then purified accord-ing to its ability to interact specifically with GCCn
Trang 11 Department of Biochemistry, Institute of Experimental Medicine, Russian Academy of Medical Sciences, St Petersburg, Russia
2 Department of Embryology, St Petersburg State University, Russia
3 Institute of Cytology, Russian Academy of Sciences, St Petersburg, Russia
4 Department of Molecular Basis of Medicine, Institute of Gene Biology, Russian Academy of Sciences, Moscow, Russia
A variety of human diseases have been associated with
expansion of CGG⁄ CCG triplet repeat tracts within
the human genome [1–3] To date, five human
chromo-somal folate-sensitive fragile sites associated with
expansion of 5¢-(CGG)n-3¢ trinucleotide repeats have
been characterized at the molecular level: FRAXA,
FRAXE, FRAXF, FRA16A, and FRA11B Expansion
at the FRAXA locus results in fragile X-mental
retar-dation syndrome, whereas expansion at the FRAXE
locus leads to mild mental retardation, and the FRA11B locus has been implicated in Jacobsen’s syndrome [3,4] In the case of the FRAXA locus, the (GCC)n triplet repeat amplification has occurred within the 5¢-UTR of the fragile X-mental retarda-tion 1 (FMR1) gene [5] Expansions of n > 200 (n < 70 is normal) followed by cytosine methylation inactivate the FMR1 gene The FMR1 gene product
is an RNA-binding protein that associates with
Keywords
FMR1; fragile X syndrome; (GCC) n ; triplet
repeats; ZF5; zinc finger transcription factors
Correspondence
S V Orlov, Department of Biochemistry,
Institute of Experimental Medicine, Russian
Academy of Medical Sciences, 197376,
Acad Pavlov Street 12, St Petersburg,
Russia
Fax: +7 812 234 0310
Tel: +7 812 346 0644
E-mail: serge@iem.sp.ru
Present address
*Department of Molecular Neurobiochemistry,
Ruhr-University, Bochum 44801,
Germany
(Received 1 May 2007, revised 20 July
2007, accepted 24 July 2007)
doi:10.1111/j.1742-4658.2007.06006.x
A series of relatively short (GCC)ntriplet repeats (n¼ 3–30) located within regulatory regions of many mammalian genes may be considered as puta-tive cis-acting transcriptional elements (GCC-elements) Fragile X-mental retardation syndrome is caused by an expansion of (GCC)n triplet repeats within the 5¢-untranslated region of the human fragile X-mental retarda-tion 1 (FMR1) gene The present study aimed to characterize a novel human (GCC)n-binding protein and investigate its possible role in the regu-lation of the FMR1 gene A novel human (GCC)n-binding protein, p56, was isolated and identified as a Kru¨ppel-like transcription factor, ZF5, by MALDI-TOF analysis The capacity of ZF5 to specifically interact with (GCC)ntriplet repeats was confirmed by the electrophoretic mobility shift assay with purified recombinant ZF5 protein In cotransfection experi-ments, ZF5 overexpression repressed activity of the GCC-element contain-ing mouse ribosomal protein L32 gene promoter Moreover, RNA interference assay results showed that endogenous ZF5 acts as a repressor
of the human FMR1 gene Thus, these data identify a new class of ZF5 targets, a subset of genes containing GCC-elements in their regulatory regions, and raise the question of whether transcription factor ZF5 is impli-cated in the pathogenesis of fragile X syndrome
Abbreviations
CAST, cyclic amplification and selection of targets; EMSA, electrophoretic mobility shift assay; GAPDH, glyceraldehyde-3-phosphate dehydrogenase; NFjB, nuclear factor kappa B; siRNA, small interfering RNA.
Trang 2polyribosomes as part of large messenger
ribonucleo-protein complex, modulating the translation of its
RNA ligands [6] FMR1 protein is thought to be
involved in RNA interference machinery [7–9]
The molecular mechanisms responsible for the
insta-bility of (GCC)n triplet repeats remain largely
unknown Several models have been suggested,
includ-ing DNA polymerase slippage [10–12], formation of
unusual secondary structures by contiguous (GCC)n
tracts [13,14], and interactions of (GCC)n triplet
repeats with DNA-binding proteins [15,16] (GCC)n
triplet repeats are not restricted to folate-sensitive
frag-ile sites in mammalian genomes [4,17] There are
rela-tively short (GCC)n triplet repeats (n¼ 3–30) located
within regulatory regions of many mammalian genes
that may operate as cis-acting transcriptional elements
(GCC-elements) [17] We previously characterized
GCC-elements within the promoter of the gene that
encodes mouse ribosomal protein L32 (rpL32) [18–20]
and within the 5¢-UTR of the human very low density
lipoprotein receptor gene (VLDLR) [21] Such
GCC-elements are transcriptionally active and can interact
specifically with human and mouse nuclear proteins
To date, only one (GCC)n-binding protein, p20, has
been characterized [16,18,22] The p20 protein was first
detected in HeLa cells as a protein interacting with the
GCC-element of rpL32 [18], and then purified
accord-ing to its ability to interact specifically with (GCC)n
triplet repeats within the 5¢-UTR of the human FMR1
gene; the corresponding gene was cloned [16,22]
Puri-fied p20 (20 kDa) was found not to be homologous to
any known DNA-binding proteins and was designated
CGG triplet repeat binding protein 1 (CGGBP1)
Sub-sequently, in cotransfection experiments, CGGBP1 has
been demonstrated to repress the activity of the FMR1
[23], VLDLR [21], and rpL32 (K B
Kuteykin-Teplya-kov, E B Dizhe, S V Orlov & A P PerevozchiKuteykin-Teplya-kov,
unpublished results) genes Thus, DNA-binding
pro-teins that interact with (GCC)n repeats may be
involved in stabilization⁄ destabilization of the triplet
repeats [16,22] and⁄ or transcriptional regulation of
GCC-triplet containing genes [19–21] Hence, in our
previous studies, we proposed the existence of other
yet uncharacterized mammalian (GCC)n-binding
pro-teins [19–21] In the present study, we report the
iso-lation of a novel (GCC)n-binding protein p56 from
human hepatoma HepG2 cells that interacts
specifi-cally with the GCC-elements of the rpL32 and FMR1
genes This protein was identified herein to be a
Kru¨ppel-like Zn-finger transcription factor ZF5
Mammalian ZF5 protein was first identified in
mouse as a transcriptional repressor of the murine
c-myc gene with a molecular mass of approximately
52 kDa It was subsequently shown to repress the her-pes simplex virus thymidine kinase gene promoter [24] ZF5 homologues have since been cloned in humans and in chicken [25–27] Vertebrate ZF5 proteins are highly conserved and contain five Kru¨ppel-like Zn-fin-gers at the C-terminus They also have a hydrophobic BTB⁄ POZ domain (124 N-terminal amino acid resi-dues), which is responsible for protein–protein inter-actions, and a nuclear localization signal [25,26] The DNA-binding domain of ZF5 consists of the third and forth Zn-fingers [28] Cyclic amplification and selection
of targets (CAST) assays performed independently by two groups confirmed the GC-rich content of ZF5 binding sites with a common core sequence 5¢-GCGCG-3¢ [28,29] Interestingly, none of the ZF5 sequences isolated by CAST assays contained (GCC)n triplet repeats
Several novel ZF5 targets have been described [29,30] Transcription factor ZF5 has been reported to have both positive and negative effects on target gene transcription Moreover, in silico content analysis of the core promoter regions of human genes revealed the presence of ZF5 consensus sites within approximately 60% of human gene core promoters [31] In spite of significant over-representation of these sites in pro-moter regions relative to nonpropro-moter background data, these results must be verified experimentally, especially in terms of the relative low complexity of the ZF5 consensus sequence In the present study, we have shown for the first time that endogenous ZF5 protein acts as a repressor of the FMR1 gene in HepG2 cells
Results
Characterization of human nuclear proteins that interact with a composite cis-acting element of rpL32 ()24 +11)
We have previously shown that the mouse rpL32 gene contains a composite cis-acting element that spans the transcription start site ()24 +11) [19]; this composite element consists of a GCC-element and a pyrimidine block (Fig 1A) Prior tissue distribution studies of mammalian nuclear proteins interacting with the rpL32 fragment ()24 +11) revealed similarities between the DNA–protein complexes formed by the rpL32 fragment ()24 +11) with nuclear proteins from human embryonic fibroblasts and HepG2 cells [19] Electrophoretic mobility shift assay (EMSA) com-petition experiments in the present study indicated that the complete rpl32 fragment ()24 +11) has a greater affinity to nuclear proteins from human fibroblasts
Trang 3than does either the GCC-element or pyrimidine block
alone (Fig 1B,C) Similar results were obtained with
nuclear proteins from HepG2 cells (data not shown)
These results suggest that this rpL32 fragment is a
syn-ergetic composite cis-acting element
EMSA was conducted with 1,10-phenanthroline (a
chelator that specifically removes Zn2+ from
metallo-proteins) [32] to examine whether Zn-finger
DNA-bind-ing human proteins are involved in DNA–protein
complexes formed by the rpL32 fragment The presence
of 1,10-phenanthroline in the reaction mixture inhibited
formation of DNA–protein complexes (Fig 1D) and the addition of Zn2+partially restored complex formation The relatively weak Zn-chelator EGTA produced less inhibition of DNA–protein interactions than 1,10-phe-nanthroline These results suggest that the DNA-binding activity of human nuclear proteins interacting with rpL32 fragment ()24 +11) depends on Zn2+, which is consistent with the supposition that those proteins con-tain Zn-finger DNA-binding domains
Although there are similarities between the DNA– protein complexes formed by the rpL32 fragment
Fig 1 Characterization of human nuclear proteins interacting with the rpL32 fragment ( )24 +11) (A) Composite cis-acting ele-ment of mouse ribosomal protein L32 gene The numbers indicate position relative to the rpL32 transcription start point The sequences corresponding to the GCC-ele-ment and pyrimidine tract are underlined (B,C) Results of EMSA competition experi-ments with nuclear proteins from human embryonic fibroblasts The competitors and their molar excesses are shown above the lanes: (GCC)3-element; rpL32 ( )6 +11)-pyrimidine part of the rpL32 fragment ( )24 +11); K – -negative control (without nuclear proteins); K+-binding reaction with
no competitors; CI and CII refer to the spe-cific DNA-protein complexes I and II formed
by the rpL32 fragment ( )24 +11); F refers
to the free rpL32 fragment ( )24 +11) (D) EMSA experiments with chelators specific
to bivalent cations: 1, free rpL32 fragment ( )24 +11) without human embryonic fibro-blast nuclear proteins (negative control);
2, binding reaction of the rpL32 fragment ( )24 +11) with human embryonic fibro-blast nuclear proteins with no chelators;
3, binding reaction with addition of 5 m M
1,10-phenanthroline; 4, binding reaction with addition of 5 m M 1,10-phenanthroline and
5 m M ZnCl 2 ; 5, binding reaction with addi-tion of 5 m M EGTA CI and CII refer to the specific DNA-protein complexes I and II formed by the rpL32 fragment ( )24 +11) (E) Southwestern assay of HepG2 nuclear proteins with the rpL32 fragment ( )24 +11) revealing the p56 and p68 bands (arrowheads).
Trang 4()24 +11) with human fibroblast nuclear proteins
and those formed with HepG2 cell nuclear proteins,
the latter were found to be much more abundant
Therefore, nuclear proteins from HeG2 cells were
employed in the subsequent Southwestern assays Two
major putative (GCC)n-binding proteins, p56 and p68,
were detected by Southwestern assays (Fig 1E)
Purification of human nuclear proteins that
interact with a composite cis-acting element of
rpL32 ()24 +11)
We designed a four-step purification procedure to
iso-late proteins that interact with the rpL32 fragment
()24 +11) (Fig 2A) This approach included a
con-sequent preparative EMSA stage, two-step purification
by DNA-affinity chromatography with the rpL32
frag-ment ()24 +11) immobilized on magnetic beads as a
substrate, and DNA-affinity chromatography on a
nonspecific substrate to remove the remaining
nonspe-cific DNA-binding proteins Figure 2B shows the
elu-tion profile of the preparative EMSA of the rpL32
fragment ()24 +11) with nuclear proteins from
HepG2 cells The first peak corresponds to unbound
labeled DNA probe, the second and third peaks
corre-spond to DNA–protein complexes formed by the
rpL32 fragment ()24 +11) with studied proteins
Fractions 21–27, corresponding to latter two peaks,
were collected, concentrated by ultrafiltration and
checked for DNA-binding activity by analytical EMSA
(Fig 2C) The slower migrating complex (Complex I)
spontaneously converted into the faster migrating one
(Complex II) (Fig 2C) Therefore, fractions 22–25
with high DNA-binding activity were pooled, analyzed
by SDS electrophoresis (Fig 2E, lane 1), and subjected
to purification by DNA-affinity chromatography
In the final purification step, the proteins were eluted
from immobilized rpL32 fragment ()24 +11) onto
immobilized human lactoferrin binding sites
(nonspe-cific DNA substrate) The DNA-binding activity
analy-sis results are summarized in Fig 3D The final fraction
containing unbound last stage proteins was analyzed by
SDS electrophoresis and compared with proteins bound
to nonspecific sorbent (negative control) (Fig 2E, lanes
3 and 2, respectively) Only two proteins, p56
(approxi-mately 56 kDa) and p68 (approxi(approxi-mately 68 kDa), that
were not in the negative control fraction remained in a
final fraction There were also two bands at 65 kDa and
72 kDa that were detected in both the negative control
and final fractions The p56 and p68 proteins in the
final fraction fit with the rpL32 fragment
()24 +11)-binding proteins detected by Southwestern assay (see
above) Together, these findings indicate that p56 and
p68 are DNA-binding proteins that interact specifically with composite cis-acting element of rpL32
Identification of rpL32-binding protein p56 by MALDI-TOF assay as a Kru¨ppel-like transcription factor ZF5
MALDI-TOF assays were employed to identify the p56 protein After SDS electrophoresis, the corres-ponding bands from silver-stained gel were cut out, digested by trypsin and subjected to analysis A typical representative spectrum fragment containing several peaks corresponding to p56-derived peptides is shown
in Figure 3A Identification of p56 protein was per-formed using the Mascot search engine Despite the presence of substantial noise, we were able to detect among the potential candidates one Zn-finger protein with a molecular weight of approximately 52 kDa: human Kru¨ppel-like Zn-finger transcription factor ZF5 (Fig 3B) The cDNA gene encoding human ZF5 protein has been cloned by two-step RT-PCR and subcloned into a pBluescript vector Experiments to identify the p68 protein are currently in progress
Recombinant fusion protein GST-ZF5-ZF5 purified from bacterial cells specifically interacts with (GCC)nrepeats
EMSA experiments with a recombinant fusion protein containing ZF5 Zn-finger domains linked to bacterial glutathione transferase (GST-ZF5) were performed to test the capacity of ZF5 to interact with the rpL32 fragment ()24 +11) and to examine the possible (GCC)n-binding activity of ZF5 The recombinant pro-tein was isolated from bacterial cells by glutathione affinity chromatography (Fig 4) Recombinant GST-ZF5 protein efficiently recognized the GST-ZF5 consensus sequence (Fig 4A, lanes 2 and 6) Molar excesses of unlabeled ZF5 binding site depleted the corresponding bands in a dose-dependent manner (Fig 4A, lanes 3–5), indicating a good specificity of the DNA–protein interactions Interestingly, the unlabeled rpL32 frag-ment ()24 +11) depleted the corresponding band even more efficiently (Fig 4A, lanes 7–9)
Similar results were obtained in reciprocal EMSA experiments (Fig 4B) Recombinant GST-ZF5 pro-tein specifically interacted with the rpL32 frag-ment ()24 +11) Moreover, the rpL32 fragfrag-ment ()24 +11) was a more effective competitor than the ZF5 binding site (Fig 4B, lanes 2–4 and 6–8, respec-tively) The irrelevant lactoferrin binding site (negative control) did not disrupt the corresponding DNA–pro-tein complex (Fig 4B, lanes 10–12) Therefore, these
Trang 5Fig 2 Purification of HepG2 nuclear proteins interacting with the rpL32 fragment ( )24 +11) (A) Summary of purification scheme (B) Pre-parative EMSA elution profile (C) Testing of the rpL32 fragment ( )24 +11)-binding activity of 21st to 27th fractions obtained on a prepara-tive EMSA purification stage (D) EMSA testing of the rpL32 fragment ( )24 +11)-binding activity in the samples after DNA affinity chromatography: 1, negative control (without proteins); 2, positive control (EMSA with HepG2 crude nuclear extracts); 3, the proteins eluted from the immobilized rpL32 ( )24 +11) fragment (first round of DNA affinity chromatography); 4, the proteins eluted after second round of DNA affinity chromatography; 5, proteins not bound to immobilized lactoferrin binding site (final fraction) (E) Analysis of fractions obtained from the preparative EMSA and DNA affinity chromatography stages by SDS electrophoresis: 1, pooled fractions 22–25 obtained from the preparative EMSA stage; 2, the proteins bound by the lactofferin binding site in the last purification stage; 3, unbound proteins obtained from the last stage of purification (final fraction) CI and CII, DNA-protein complexes I and II, respectively C¢, Nonspecific DNA–protein complexes;
F, free rpL32 fragment ( )24 +11), p56 and p68- purified proteins specifically interacting with the rpL32 fragment ()24 +11).
Trang 6Fig 3 Identification of p56 polypeptide as Kru¨ppel-like transcription factor ZF5 by MALDI-TOF assay (A) Fragment of MALDI-TOF spectrum of p56-derived peptides (B) Mascot search results Peptides with matched mass values are listed with their locations in the total ZF5 protein.
Trang 7data provide a robust demonstration of the capacity of
ZF5 to interact specifically with rpL32 composite
cis-acting element Furthermore, the affinity of
recombi-nant GST-ZF5 for the rpL32 fragment ()24 +11) is
higher than its affinity for a classic ZF5 binding site
EMSA experiments with a labeled GCC-element
probe or pyrimidine motif probe were conducted to
determine whether the ZF5 binds to the GCC-element
or pyrimidine motif of rpL32 composite cis-acting
ele-ment The pyrimidine part of the rpL32 fragment
bound nuclear proteins from HepG2 cells, but did not interact with recombinant GST-ZF5 (Fig 4C) Mean-while, a (GCC)9 sequence corresponding to a 5¢-UTR fragment of the human FMR1 gene (normal allele) bound specifically to GST-ZF5 Molar excesses
of unlabeled rpL32 fragment ()24 +11), (GCC)9 sequence, or ZF5 binding site, but not lactoferrin bind-ing site, efficiently disrupted (GCC)9⁄ ZF5 complexes (Fig 4D,E) These data demonstrate unequivocally that ZF5 transcription factor interacts specifically with
Fig 4 Interactions of recombinant GST-ZF5 with GCC-element of the rpL32 promoter (A) EMSA using the ZF5 binding site as a probe (B) EMSA using the rpL32 fragment ( )24 +11) as a probe (C) EMSA using the pyrimidine site of rpL32 ()6 +11) as a probe (D,E) EMSA using a (GCC)9-element as a probe K – , without any proteins; K + , EMSA with purified recombinant GST-ZF5 without competition; HepG2, EMSA with nuclear proteins from HepG2 cells Competitors are shown above each image Triangles above images represent the increasing amounts of competitor applied C, Specific DNA–protein complex; F, free DNA probe.
Trang 8(GCC)n triplet repeats and, hence, may be considered
a novel (GCC)n-binding protein
ZF5 overexpression leads to repression of rpL32
promoter that contains GCC-element
The capacity of ZF5 transcription factor to interact
with GCC-elements suggests that it may regulate rpL32
gene expression To investigate this possibility, we
con-structed a eukaryotic expression vector containing the
ZF5 coding region under the control of the human
cytomegalovirus early gene promoter (pCMVZF5) and
performed cotransfection experiments When HepG2
cells were transiently cotransfected with a pCMVZF5
expression vector and pL32luc plasmid containing a
firefly luciferase reporter gene under the control of
rpL32 promoter, ZF5 overexpression strongly
down-regulated rpL32 promoter activity in dose-dependent
manner (Fig 5A) relative to its impact on the activity
of a synthetic minimal promoter containing five tandem
cis-acting elements for transcription factor nuclear
fac-tor kappa B NFeˆB (negative control) (Fig 5B) These
results demonstrate that human ZF5 can regulate
tran-scription of genes that contain (GCC)n triplet repeats
within their regulatory regions
Endogenous ZF5 down-regulates expression of
the FMR1 gene in HepG2 cells
RNA interference was used to test whether ZF5 is
involved in regulation of the human FMR1 gene
Three small interfering RNAs (siRNAs) matching
dif-ferent regions of ZF5 mRNA were selected based on
general rules for siRNA design [43] The efficiency of
ZF5 down-regulation in HepG2 cells by siRNA
trans-fection was estimated by semiquantitative RT-PCR
HepG2 cells were transfected with the most active
siRNA (matched bases 945–963 of human ZF5
mRNA; accession number D89859) The amount of
FMR1 mRNA in siRNAZF5 transfected cells
har-vested after 48 h was accessed by semiquantitative
RT-PCR and compared with the mRNA from
non-transfected cells and cells non-transfected with control
siRNA that was not homologous to any human
mRNAs (Fig 6A,B) ZF5 down-regulation increased
FMR1 mRNA levels in HepG2 cells; transfection of
HepG2 cells with control siRNA did not influence
FMR1 mRNA levels It was suggested previously that
FMR1 protein may be involved in the RNA
interfer-ence machinery [7–9] Thus, accumulation of FMR1
mRNA may be due to stimulation of FMR1 gene
expression in response to activation of RNA
interfer-ence machinery in the cells trasfected by siRNAZF5
but not to down-regulation of ZF5 To test this assumption, we transfected HepG2 cells by siRNA for human grp58 gene [44] siRNAgrp58 down-regu-lated expression of grp58 gene but did not affect FMR1 mRNA (Fig 6C) This, these data indicate that endogenous transcription factor ZF5 is a repres-sor of the human FMR1 gene
Discussion
CGGBP1 is the only DNA-binding protein described
to date that specifically interacts with (GCC)n repeats [16,22] Two observations led us to hypothesize that other (GCC)n-binding proteins exist and to search for
Fig 5 Effect of ZF5 overexpression on activity of the rpL32 pro-moter relative to the synthetic NFjB-dependent propro-moter (A) HepG2 cells were cotransfected with pL32luc (5 lg), pCMVL (2.5 lg) and different amounts of pCMVZF5: 1, without pCMVZF5;
2, with 50 ng pCMVZF5; 3, with 150 ng pCMVZF5; 4, with 375 ng pCMVZF5 (B) HepG2 cells were cotransfected with pNFjBluc (5 lg), pCMVL (2.5 lg) and different amounts of pCMVZF5: 1, with-out pCMVZF5; 2, with 50 ng pCMVZF5; 3, with 150 ng pCMVZF5;
4, with 375 ng pCMVZF5.
Trang 9possible novel human (GCC)n-binding transcription
factors First, the DNA-binding activity of nuclear
proteins that interact with the rpL32 fragment
()24 +11) is Zn2+-dependent (Fig 1D), although
the CGGBP1 protein does not contain any
Zn-bind-ing domains [22] Second, the affinity of purified
CGGBP1 to (GCC)n repeat sequences is dramatically
decreased if there are fewer than eight repeats [16]
However, a few GCC repeats are sufficient to
regulate an artificial herpes viral thymidine kinase
nuclear proteins with high affinity [15,16] However the regulatory regions of many mammalian genes contain shorter (GCC)nmotifs (n¼ 3–10) that are adjacent to binding sites of other transcription factors (i.e Sp1, Egr1, WT1) [17] Thus, low affinity binding of short (GCC)n repeats to (GCC)n-binding proteins may be compensated for by cooperative protein–protein inter-actions with transcription factors that bind to adjacent sequences
We previously characterized a fragment of mouse rpL32 promoter ()24 +11) containing an adjacent GCC-element ()19 )6) and pyrimidine motif ()5 +11) as a composite cis-acting element that interacts with nuclear proteins from mammalian cells [18–20] The present data (Fig 1B,C) suggest that there may be cooperation between DNA–protein and protein–protein interactions within complexes formed
by the rpL32 fragment ()24 +11) We proceeded to detect, by Southwestern assay, only two polypeptides
in nuclear extracts from HepG2 cells that interact with the rpL32 fragment ()24 +11) (Fig 1E) These find-ings indicate that the rpL32 fragment ()24 +11) would serve as an effective bait for purifying GCC-ele-ment binding proteins
The rpL32 fragment ()24 +11) formed two major complexes with nuclear proteins from HepG2 cells and human embryonic fibroblast cells (Fig 1B,C) In our attempt to isolate these complexes, an exploratory pre-parative EMSA revealed that the slower migrating Complex I converted spontaneously to the faster migrating Complex II (Fig 2C) This observation sug-gests that the difference between the DNA-binding complexes might be caused by reversible post-transla-tional modifications of rpL32-binding proteins Alter-natively, those complexes may differ by unstable proteins that were lost during the purification process Additional experiments are needed to distinguish between these possibilities
We detected two purified DNA-binding proteins spe-cifically interacting with the rpL32 fragment ()24 + 11) that coincided with the rpL32-binding proteins revealed by Southwestern assay The results of the MALDI-TOF analysis indicated that one of these, p56,
is a Kru¨ppel-like transcription factor ZF5 Although all ZF5-binding sites found previously by CAST assay are GC-rich, none are within (GCC)ntriplet repeats [28,29] Our EMSA experiments using recombinant protein
Fig 6 Down-regulation of endogenous ZF5 in HepG2 cells by RNA
interference (A) Effects of HepG2 cell transfection by siRNAs on
ZF5 expression (semiquantitative RT-PCR): lanes 1–6, RT-PCR of
GAPDH mRNA (18 cycles, control); lane 7, marker; lanes 8–13,
RT-PCR of ZF5 mRNA (27 cycles); 1, 8, negative control; 2, 9,
trans-fection of HepG2 cells by siRNA ZF5 with oligofectamine; 3, 10,
transfection of HepG2 cells by control siRNA with oligofectamine;
4, 11, transfection of HepG2 cells by siRNA ZF5 with RNAFect; 5,
12, transfection of HepG2 cells by control siRNA with RNAFect; 6,
13, untreated cells (B) Effects of HepG2 cell transfection by
siRNA ZF5 on FMR1 expression (semiquantitative RT-PCR, 34 cycles):
1, marker; 2, negative control; 3, transfection by siRNA ZF5 with
oli-gofectamine; 4, transfection by control siRNA with olioli-gofectamine;
5, transfection by siRNAZF5with RNAFect; 6, transfection by control
siRNA with RNAFect; 7, untreated cells (C) Effects of HepG2 cell
transfection by siRNAgrp58 on FMR1 expression (semiquantitative
RT-PCR): 1–3, RT-PCR of GAPDH mRNA (18 cycles, control); 4,
mar-ker; lanes 5–7, RT-PCR of grp58 mRNA (27 cycles); 8–10, RT-PCR of
FMR1mRNA (34 cycles); 1, 5, 8, negative control; 3, 7, 10, untreated
cells; 2, 6, 9, transfection by siRNAgrp58with RNAFect.
Trang 10containing the DNA-binding domain of ZF5 fused with
GST confirmed that recombinant GST-ZF5 specifically
interacts with its canonical binding site, the rpL32
fragment ()24 +11), and the GCC-element alone
(Fig 4) Moreover, competition experiments showed
that the affinity of ZF5 for the rpL32 fragment
()24 +11) is greater than its affinity for its own
consensus sequence
Although the rpL32 fragment ()24 +11) contains
a core part of the ZF5 consensus sequence GCGC
immediately before the GCC-element (Fig 1A), a
(GCC)9 sequence that does not contain any ZF5
consensus core was recognized specifically by ZF5
(Fig 4C) Thus, the high affinity of ZF5 for the rpL32
fragment may be due to cooperation interactions of
ZF5 with both its consensus core sequence and the
GCC-element Hence, ZF5 appears to bind
GCC-ele-ments with an affinity that is similar to that for its
consensus sequence This finding may explain why
pre-vious CAST assays did not reveal GCC-elements to be
among ZF5-binding sites The affinity of ZF5–DNA
interactions may depend on the length of the binding
site Random sequences used for CAST assays are
approximately 20 bp in length [28], and the selection
procedure is based on single site affinities to the
protein of interest In the case of the rpL32 fragment
or the (GCC)9 sequence, there are several repeated
ZF5-binding sites within a single oligonucleotide
Therefore, the affinity of those sequences for ZF5
revealed in the present study may be the result of
cooperative interactions between several ZF5 molecules
and adjacent ZF5-binding sites Such a cooperative
mechanism is especially interesting with respect to the
role of (GCC)n triplet repeat amplification in
inactiva-tion of the FMR1 gene That is, ZF5 affinity for
(GCC)n repeat tracts in the FMR1 gene would be
expected to be greatly enhanced in alleles with very
extended (GCC)nrepeats
Mammalian ZF5 protein was first identified as a
transcriptional repressor of the murine c-myc gene
Although ZF5-mediated transcriptional activation has
been reported [29,30], ZF5 is most often observed to
down-regulate target genes Here, we show that ZF5
overexpression led to down-regulation of the rpL32
promoter Moreover, down-regulation of ZF5 by
RNA interference caused up-regulation of the FMR1
gene These data clearly support the assertion that
endogenous ZF5 acts as a transcriptional repressor of
the human FMR1 gene Given that FMR1 inactivation
in FMR patients has been attributed to long (GCC)n
triplet repeats within the 5¢-UTR of the FMR1 gene,
further investigations to clarify the role of ZF5 in
FMR pathogenesis are warranted
In conclusion, we report that p56 protein purified from HepG2 cells interacts with GCC-elements of the rpL32 promoter and the 5¢-UTR of the human FMR1 gene We further identify p56 as the Kru¨ppel-like transcription factor ZF5 We show, for the first time, that recombinant mammalian ZF5 protein interacts specifically with (GCC)n triplet repeats, and further show that endogenous ZF5 acts to down-reg-ulate the FMR1 gene The present study has revealed
a novel class of ZF5 target genes that have GCC-ele-ments within their regulatory regions and implicates ZF5 in the pathogenesis of fragile X-mental retarda-tion syndrome
Experimental procedures
Materials
Chemicals were purchased from Sigma (St Louis, MO, USA), Amersham Biosciences (Piscataway, NJ, USA), Roche Applied Science (Mannheim, Germany), Invitrogen (Carlsbad, CA, USA), Promega (San Luis Obispo, CA, USA) and from local Russian manufacturers (analytical or high purity grade) Enzymes used in gene engineering were purchased from Fermentas (Vilnius, Lithuania) and the Sci-entific Industrial Corp SibEnzyme (Novosibirsk, Russia)
Genetic constructions
pL32luc plasmid containing the reporter gene encoding firefly luciferase under the control of rpL32 promoter ()155 +195 relative to the transcription start point) was constructed as follows First, rpL32 promoter was cut out from pL3A plasmid containing the genomic rpL32 gene (a gift from Dr N V Tomilin, Institute of Cytology, Russian Academy of Sciences) by AccI and SmaI and inserted into pUC19 vector Next, it was digested from pUC19 by EcoRI, blunt ended by Klenow fragment of Escherichia coli DNA polymerase I, digested by HindIII and inserted into a pGL3basic plasmid containing the luciferase gene (Pro-mega) The pCMVL plasmid containing the reporter gene lacZ and encoding bacterial b-galactosidase driven by a promoter of early human CMV genes have been described previously [19] pNFjBluc plasmid contains luciferase gene driven by synthetic minimal promoter carrying five binding sites for transcription factor NFjB was purchased from Stratagene (La Jolla, CA, USA)
Cell culture
HepG2 cells were obtained from the Cell Culture Bank of the Institute of Cytology, Russian Academy of Sciences Human embryonic fibroblast cells were obtained from the Institute of Influenza, Russian Academy of Medical