Dynamic interaction of HMGA1a proteins with chromatin

The functional activity of the genome is controlled by its packaging into chromatin. Recent in vivo imaging approaches revealed the dynamic behavior of structural chromatin proteins, such as HMGN proteins and histone H1 (Lever et al., 2000; Misteli et al., 2000; Phair and Misteli, 2000) or HP1 (Cheutin et al., 2003). Their kinetic properties were summarized in a `stop and go' model, in which the molecules bind transiently to chromatin, and diffuse through the nucleoplasm until they find another binding site (Misteli, 2001; Misteli et al., 2000). The residence times for histone H1 were found to be 3-4 minutes, compared with seconds in the case of HMGN proteins. Upon chromatin hyperacetylation, the residence times were reduced indicating their dependence on the specific properties of the chromatin as well, as the functional status of the proteins itself (Misteli, 2001; Misteli et al., 2000). The striking mobility of chromatin proteins has been mechanistically implicated in local and global reorganizations of chromatin, that is, the rapid association and dissociation of chromatin proteins provide free binding sites for the same or other chromatin components, which could alter the chromatin status (Misteli, 2001). Indeed, it was shown that HMGN proteins decrease the residence time of H1 on chromatin and, therefore, may counteract the inhibitory effects of the H1-induced higher-order chromatin structure (Catez et al., 2002).

HMGA1^† proteins preferentially bind to the minor groove of AT-rich B-DNA with three AT-hook binding motifs (Reeves and Nissen, 1990). As shown by immunocytological approaches, HMGA1a/b proteins preferentially localize to the heterochromatin mass (Amirand et al., 1998; Martelli et al., 1998). They bind to DNA elements termed scaffold attachment regions (Zhao et al., 1993) and have been proposed to function as competitors of H1-mediated general repression of transcription in vitro (Käs et al., 1993; Zhao et al., 1993). HMGA1 also has the ability to bind to DNA packaged in nucleosomes and this ability is modulated by posttranslational modifications (Banks et al., 2000; Reeves, 2001; Reeves et al., 2000). Furthermore, HMGA1 proteins are involved in regulating the expression of specific genes, reviewed elsewhere (Reeves and Beckerbauer, 2001). The most accepted model of how they function in gene regulation is through either the facilitation, or inhibition, of the formation of `enhanceosomes' (Thanos and Maniatis, 1995).

HMGA gene expression is maximal during embryonic development (Chiappetta et al., 1996), drops off in most adult tissues and is low, or undetectable, in fully differentiated or non-dividing adult cells (Bustin and Reeves, 1996; Lundberg et al., 1989). Overexpression of HMGA proteins correlates with neoplastic transformation and tumor progression in many malign tumors (Tallini and Dal Cin, 1999). In most benign mesenchymal neoplasias, chromosomal translocations lead to fusion proteins with multiple AT-hooks (Hess and Kossev, 2002).

To gain insight into the in vivo distribution, function and dynamics of HMGA1a, we expressed GFP fusion proteins in HepG2 cells. First, the distribution and chromatin binding of the chimeric proteins was investigated by immunolocalization and extraction experiments. Second, we used fluorescence recovery after photobleaching (FRAP) to examine the dynamic behavior of the fusion proteins over euchromatin, heterochromatin and on mitotic chromosomes. These experiments revealed that HMGA1a proteins belong to the highly mobile components of interphase and mitotic chromatin, although with different kinetic properties. Third, we used a set of point-mutated fusion proteins and FRAP to analyze the functional role of the AT-hook-binding motifs. We identified the first two of the three AT hooks as main players mediating DNA/chromatin binding. Fourth, the contribution of posttranslational modifications on HMGA1a dynamics in vivo was analyzed by FRAP after inhibition of kinase and histone deacetylase activities. These experiments showed that the activity of p^34/cdc2 kinase and protein kinase C as well as histone deacetylases regulate the nuclear dynamic equilibrium of HMGA1a. Fifth, we analyzed the contribution of specific phosphorylation sites on the dynamic mobility of HMGA1a in vivo. The threonines in front of each AT hook were mutated into alanines to yield single-, double- and triple-point-mutated fusion proteins. FRAP experiments were performed to compare the mobility of the chimeric proteins in heterochromatin and on chromosomes. Interestingly, these data suggested that, compared with euchromatin, the increased residence time in heterochromatin and on mitotic chromosomes correlates with an increased phosphorylation.

The coding sequence of human HMGA1a was PCR amplified out of 2 μl reverse-transcribed HepG2 RNA using the primers 5′ ATGAGTGAGTCGAGCTCGAAGTCCAGC 3′ and 5′ ACTGCTCCTCCTCCGAGGACTCC 3′ (Interactiva). This and all other PCR products were subcloned by TOPO™TA cloning according the manufacturers instructions (Invitrogen). The HMGA1 cDNA then was EcoRI inserted into or pEGFPN1 (Clontech).

Site-directed mutagenesis of human HMGA1a was performed essentially as described in the QuickChange® Site-Directed Mutagenesis Kit protocol (Stratagene). For mutation of hHMGA1a phosphorylation sites, alanine codon substitutions were introduced at the nucleotides coding for the amino acid threonine at positions 21 (T21A), 53 (T53A) and 78 (T78A) of the hHMGA1a protein sequence as reported elsewhere (Siino et al., 1995).

AT-Hook point mutations were introduced by glycine substitution of the arginine codon following the central glycine codon of the conserved AT-hook motif PRGRP at positions 28 (R28G), 60 (R60G) and 86 (R86G). The following primers (Thermo Hybaid) were used in the PCR reactions, respectively: R28G 5′ CGGGGCCGGGGCGGGCCGCGCAAGCAG 3′; R28G reverse 5′ CTGCTTGCGCGGCCCGCCCCGGCCCCG 3′; R60G 5′ CCTAAGAGACCTCGGGGCGGACCAAAGGGAAGC 3′; R60G reverse 5′ GCTTCCCTTTGGTCCGCCCCGAGGTCTCTTAGG 3′; R86G 5′ GGAAGGAAACCAAGGGGCGGACCCAAAAAACTGGAG 3′; R86G reverse 5′ CTCCAGTTTTTTGGGTCCGCCCCTTGGTTTCCTTCC 3′. Combinations of mutations were made sequentially. Derived clones were verified by sequencing.

HepG2 cells were transiently transfected using effectene (Quiagen) and 400 ng of plasmid DNA as specified by the manufacturer. Transfected cells were grown on coverslips over night at 37°C and 5% CO₂ to 50% confluence. Transfection rate was at least 50%, but routinely more than about 60%. Expression of HMGA1a-GFP did not interfere with cell growth over multiple cell cycles. For chromatin acetylation, cells were incubated with 500 ng/ml Trichostatin A (TSA, Calbiochem) for 2 hours. Roscovitine (25 μM, Calbiochem) was incubated for 4 hours. A cell-permeable-specific inhibitor for protein kinase C, myristoyl-Phe-Ala-Arg-Lys-Gly-Ala-Leu-Arg-Gln-OH (ICN Biomedicals), was used at 10 μg/ml for 4 hours.

For live-cell imaging, cells were grown on coverslips that were mounted on a slide chamber with culture medium. Cells were analyzed within the following 20-30 minutes at room temperature with a Leica TCS-SP. FRAP experiments at 37°C, 5% CO₂ showed now differences in recovery kinetics compared with those obtained at room temperature (not shown). Cells have been analyzed using the 488 nm laser line for GFP of an Ar/Kr laser (488 nm: 18 mW nominal output, pinhole setting at 1; 568 nm: 17 mW nominal output, pinhole setting at 1, ×40 neofluar objective, N.A. 1.25). For bleaching, one single scan was acquired, followed by a single bleach pulse of 1 second using a spot of approximately 1 μm in radius without imaging. For imaging, laser power was attenuated to 4% of the bleach intensity. Single-section images were then collected at 1 second intervals (20 images), 2 seconds intervals (10 images), 5 seconds intervals (10 images) and in 10 seconds intervals (20 images). To reduce chromosome movements in mitotic cells, FRAP experiments were performed in the presence of 10 μM taxol. Average fluorescence loss during imaging was 2-3%. FRAP recovery curves were generated according to Phair and Misteli (Phair and Misteli, 2000) from background subtracted pictures. The quantitative values represent averages of at least 10 cells from three independent experiments. As standards we used GFP, H3-GFP and fixed cells. Some recoveries were observed that were greater than 100% because of the contribution of electronic noise in the detector system. Student's t-test was used to determine the statistical significance of the results.

For DNA counterstaining with Hoechst 33258, transfected cells were briefly permeabilized with 0.1% Triton for 1 minute and then fixed in 2% formaldehyde in PBS for 10 minutes at room temperature. Cells then were permeabilized with 100 μl 0,1% Triton X-100 in PBS for 10 minutes at room temperature and Hoechst 33258 was added at a final concentration of 500 ng/ml. Cells were washed and mounted in Mowiol as described (Hock et al., 1998). Hoechst-stained cells have been investigated with a confocal laser scanning microscope using the 364 nm laser line of a Kr/Ar laser (Zeiss 510) or on a Zeiss Axiophot equipped with a Pixera digital system. Immunfluorescence procedure on formaldehyde fixed cells was essentially as previously described (Hock et al., 1998). Primary antibodies used for immunfluorescence experiments were anti-HMGA1 (gift from Ray Reeves), anti-sc35 (dilution 1:2000; Sigma) and anti-histone H1 (dilution 1:20; ICN). Spreading of native chromosomes was essentially as described (Christensen et al., 2002).

For in situ DNA-staining with Hoechst, transfected cells were permeabilized with 0.05% Triton X-100 in PBS containing 10 μg/ml Hoechst. Cells were incubated for 2-5 minutes and monitored immediately. Longer exposure to Hoechst led to displacement of HMGA1a-GFP.

In situ run-on transcription was performed as described previously (Hock et al., 1998). After BrUTP incorporation, cells were fixed and permeabilized as described above. Incorporated BrUTP was detected using anti-BrdUTP (Roche, Mannheim, diluted 1:25). Optical sections were recorded with a Leica TCS-SP using the ×40 neofluar oil-immersion objective (N.A. 1.25) with a pinhole setting at 0.8 and AOTF for 488 nm at 35%, and AOTF for 568 nm at 50%. Fluorescent signals of GFP and Texas-red were recorded simultaneously. Dichromatic pictures were recorded in parallel. No crosstalk was observed.

For in situ salt extraction, cells grown on coverslips were washed twice in PBS and incubated in 3 ml cytoskeleton buffer [CSK; 100 mM NaCl, 300 mM sucrose, 10 mM Pipes, pH 6.8, 3 mM MgCl₂, 0.5% Triton X-100, 1 mM PMSF, Leupetin, Pepstatin, Bestatin and RNasin (MBI)] for 10 minutes at 4°C. CSK was removed by pipetting and cells were fixed in 2% formaldehyde, or incubated further for salt extraction in 3 ml CSK with 350 mM NaCl for 5 minutes at 4°C or incubated with DNase I in digestion buffer [like CSK but with 50 mM NaCl and 20 U/ml RNasin (MBI) and 200 U/ml DNase I (Roche)] for 30 minutes at room temperature. Cells then were fixed and processed for immunfluorescence analyses as described above.

Localization of HMGA1a-GFP on ultrastructural level was perfomed essentially as described previously (Hock et al., 1998) using monoclonal antibodies directed to GFP (Roche) at 3 μg/ml and secondary antibodies coupled to 12 nm gold particles (Dianova, diluted 1:10).

To test salt extraction properties of HMGA1a-GFP fusion proteins, HepG2 cells were transfected with human HMGA1a-GFP or GFP as a control, were washed twice with PBS and then incubated with PBS containing 0.5% Triton X-100 or PBS containing 350 mM salt and 0.5% Triton X-100 for 10 minutes at room temperature. Supernatants were collected and solubilized. Proteins were precipitated with ice cold acetone. The precipitated proteins were washed twice with 70% acetone, air dried and resuspended in SDS sample buffer.

SDS-PAGE and transfer onto nitrocellulose was performed as described previously (Hock et al., 1998). Dilution of primary antibodies for immunoblots were 0.2 μg/ml for anti-GFP (Roche), 0.5 μg/ml for anti-HMGA1a (Santa Cruz), 0.3 μg/ml anti-pep2 (Hock et al., 1998). The monoclonal antibody directed against actin (Gonsior et al., 1999) was used at a dilution of 1:1000. Blocking and incubation was performed in 5% non-fat dry milk/TBS for 1 hour at room temperature except for the HMGA1 antibody, where the blocking was in TBS plus 0.1% Tween-20. Washing and detection was performed as described.

During interphase, HMGA1-GFP proteins were distributed throughout the cell nucleus excluding the nucleoli and were enriched in multiple nuclear domains or foci (Fig. 1A,a,b). The fusion proteins colocalized with their endogenous counterparts in transfected cells and in an identical pattern (Fig. 1A,c-c″). Furthermore, the distribution of the fusion proteins was comparable with that of endogenous proteins in non-transfected cells (Fig. 1A,c-c″). The HMGA1a-GFP containing foci overlapped with high local concentrations of DNA as shown by in situ Hoechst staining of unfixed and gently permeabilized HepG2 cells (Fig. 1B,a-a′) as well as in fixed HepG2 cells (Fig. 1B,b-b′). This indicated that HMGA1a-GFP preferentially localizes to heterochromatin. Essentially the same results were obtained with frog, mouse and other human cell lines (data not shown). Our finding that HMGA1a-GFP is enriched in heterochromatin is compatible with previous immunocytological data on the localization of endogenous HMGA1 proteins (Amirand et al., 1998; Martelli et al., 1998).

The preferential association of HMGA1a-GFP with heterochromatin was further supported by the overlapping distribution of HMGA1a-GFP and histone H1 (Fig. 2A,a-a″). Furthermore, in HMGA1a-GFP-expressing cells, transcription sites were labeled by in situ run-on transcription using BrUTP incorporation. Notably, the majority of HMGA1a-containing foci did not overlap with the BrUTP incorporation sites, further indicating that HMGA1a-GFP is preferentially associated with transcriptionally inactive chromatin (i.e. heterochromatin). In control experiments, when we examined the distribution of HMGN proteins, which are known to localize preferentially to transcriptionally active chromatin (Hock et al., 1998), there was a clear overlap between the sites of HMGN2-GFP proteins and that of BrUTP-labeled nascent transcripts (Fig. 2A,c-c″). Interestingly, electron microscopic (EM) immunogold localizations of HMGA1a-GFP revealed an accumulation of gold particles at electron-dense chromatin structures, most likely representing heterochromatin foci (Fig. 2B,b′). The EM-studies also showed that HMGA1a-GFP expression did not alter bulk chromatin structure at the ultrastructural level (Fig. 2B).

The extractability of the fusion proteins was investigated in situ and by western blot experiments using permeabilized transfected cells. In both assays, HMGA1a-GFP proteins were salt extractable with 350 mM salt (Fig. 3A,e-h and 3B,a), which is characteristic for members of all HMG protein families. Endogenous HMGA1a (as well as HMGN) proteins shared the same solubility properties as the fusion proteins, indicating proper DNA binding of HMGA1a-GFP (Fig. 3B,b,c). In addition, digestion of DNA by DNase I resulted in a total release of HMGA1a-GFP, indicating that the fusion proteins were not associated with non-chromatin structures (Fig. 3A,j). As a control for the in situ salt extraction and DNase I digestion experiments, we localized sc35, a non-snRNP splicing factor (Spector et al., 1991). Both treatments did not solubilize sc35, which is in striking contrast to the behavior of HMGA1a-GFP (Fig. 3A,g,h,l,m). Significantly, when cells were treated with a buffer of low ionic strength (CSK buffer), sc35 localized outside of regions with high HMGA1a concentrations (Fig. 3A,c,d).

During mitosis, HMGA1a-GFP proteins were associated with the chromosomes of live cells without indications of banding patterns or axial concentrations (Fig. 4A,a,b). When we prepared native chromosome spreads from HMGA1a-GFP-expressing cells (see Christensen et al., 2002), the homogenous distribution of HMGA1a-GFP was lost over time and resulted in a more-banded pattern after an incubation time of 1-2 hours in 75 mM KCl (Fig. 4B). In addition, after prolonged incubation of the chromosomes, HMGA1a-GFP disappeared from the peripheral regions and became enriched in the more-axial regions of the chromosomal arms (Fig. 4B,b″). Thus, HMGA1a proteins are not stably bound to mitotic chromosomes but slowly dissociate, or redistribute, upon exposure of isolated chromosomes to an isotonic buffer.

To investigate HMGA1a dynamics in living cells, we used fluorescence recovery after photobleaching (FRAP). Upon bleaching of a small nuclear area of approximately 1 μm in diameter over euchromatin or heterochromatin, recovery of fluorescence was measured (Fig. 5, Table 1). For these experiments, heterochromatin was defined on the basis of morphological criteria as areas strongly labeled with HMGA1a-GFP; euchromatin was defined morphologically as areas weakly labeled with HMGA1a-GFP.

In euchromatin, complete fluorescence recovery of the bleached spot occurred in approximately 23 seconds. After less than 1 second, 50% of the prebleach fluorescence and after 4.2 seconds already 80% of the prebleach fluorescence was regained.

In heterochromatin, fluorescence recovery was complete after approximately 37 seconds with 50% and 80% recovery of the prebleach intensity after 3.3 seconds and 12 seconds, respectively (Table 1). Thus, the mobility of HMGA1a-GFP in heterochromatin is significantly slower (P<0.004) than in euchromatin, indicating more-tightly bound HMGA1a-GFP in heterochromatin. Irrespective of these differences, the FRAP data show that HMGA1a proteins belong to the highly mobile components of chromatin. Remarkably, the recovery kinetics over heterochromatin were quite variable depending on where the bleach spot was set. This was especially striking when the recoveries to 80% of the prebleach intensity were compared. Recovery to 80% of bleached heterochromatin located directly near the nucleolus was significantly faster (7.3 seconds) as compared with heterochromatin located within the nucleoplasm (19.5 seconds; P<0.0001; Fig. 5C). Notably, kinetic differences as a function of intranuclear localization were found exclusively in the case of HMGA1a proteins and were undetectable when the dynamic properties of HMGN- or HMGB-GFP were probed. In these cases, fluorescence recovery resulted in kinetics shown in Fig. 5E and were independent from the location of the bleached region.

To apply FRAP experiments on mitotic chromosomes, positional movements of the metaphase plate had to be prevented by taxol treatment. Upon bleaching a small region of metaphase chromosomes, the fluorescence signal recovered completely within approximately 53 seconds (Fig. 5D; Table 2), which is significantly slower than the average recovery time of 37 seconds over interphase heterochromatin (P<0.005; Table 1). This observation shows that the HMGA1a proteins of mitotic chromosomes are permanently exchanged but that their residence time on mitotic chromosomes is higher compared with interphase heterochromatin.

HMGA proteins contain three DNA-binding domains termed AT hooks. In mammals, the AT-hook motifs share the consensus sequence P-R-G-R-P flanked by other positively charged residues (Reeves and Beckerbauer, 2001). To investigate how the individual AT-hook motifs I, II and III contribute to DNA binding in vivo, we have replaced the second arginine of the P-R-G-R-P consensus with glycine. The arginines were chosen because the side chains of the arginines are responsible for the contact of the AT hook with DNA (Huth et al., 1997). GFP-fusions of the point-mutated proteins were transiently expressed in human cells and their DNA-binding properties studied by FRAP. The fusion proteins examined were point mutated in the AT hook I (R28G), AT hook II (R60G) or AT hook III (R86G), double-point-mutated in AT hook I+II (R28,60G), I+III (R28,86G), II+III (R60,86G) or triple-point-mutated in AT hook I+II+III (R3×G).

Compared with wild-type HMGA1a, point-mutations of either AT hook I, II or III caused a more diffuse and less punctuate distribution in interphase cell nuclei (Fig. 6A, upper row, images R28G, R60G and R86G). However, in all three cases the mutated fusion proteins still accumulated in heterochromatin blocks. By contrast, mutation of two (Fig. 6A, images R28,60G, R28,86G and R60,86G) or all three AT-hook motifs (R3×G) caused an almost homogenous distribution of the fusion proteins throughout the entire nucleus.

In mitotic cells, mutations of either AT hook I (R28G), II (R60G) or III (R86G) induce no (R28G), or only a slight, increase of the cytoplasmic non-chromosome-bound fraction of the GFP fusion proteins (Fig. 6A, lower panel) compared with wild-type constructs (Fig. 6A, wt). The mitotic chromosomes still fluoresced, indicating that two functional AT-hook domains are sufficient to mediate HMGA1a binding to chromosomes. Expression of fusion proteins with two mutated hook domains clearly compromised DNA binding as shown by the decrease of chromosomal fluorescence and increase in cytoplasmic fluorescence (Fig. 6A). When all three hook domains were mutated (R3×G), the fusion proteins were distributed evenly without any preferential chromosomal binding (Fig. 6A, lower panel, R3×G).

To quantify the interaction of the various AT-hook mutants with DNA, we performed FRAP experiments. Here, the bleach spots were set over areas strongly labeled with mutant HMGA1a-fusion proteins located within the nucleoplasm and the recovery kinetics were compared with those obtained after bleaching similar regions labeled by the wild-type fusion protein. Mutation of either AT hook I (R28G) or II (R60G) significantly increased the mobility of the fusion proteins compared with those obtained for the wild-type protein (Fig. 6B, Table 1). Interestingly, both single-domain mutants showed a comparable increase in the kinetics of fluorescence recovery (Fig. 6B, Table 1), indicating an equal contribution of AT hook I and II for DNA binding. By contrast, the mobility of AT hook III mutants (R86G) was only moderately increased (Fig. 6B, Table 1). Thus, this domain contributes to the interaction of HMGA1a with DNA to a lesser degree than AT hook I or II.

The double domain mutants (AT hook I + II, I + III and II + III) caused an increased mobility as compared with the single domain mutants (Fig. 6B, Table 1). When we mutated all three AT-hook domains, the interaction of the mutant protein (R3×G) with DNA was even more impaired as evidenced by its relatively high mobility and full fluorescence recovery within 8 seconds compared with 12 seconds of the double mutant (Fig. 6B, Table 1). Taken together, our results demonstrate that the AT-hook motif governs the HMGA distribution in vivo and that all three AT-hook motifs contribute to the interaction of HMGA1a proteins with DNA, although to different extents.

HMGA1 proteins are among the most extensively modified nuclear proteins [e.g. phosphorylations, acetylations and methylations; reviewed in (Banks et al., 2000; Reeves and Beckerbauer, 2001)]. Protein kinase C (PKC) and p³⁴-kinase have both been implicated in the phosphorylation of HMGA1 proteins in vivo (Banks et al., 2000; Schwanbeck and Wisniewski, 1997; Xiao et al., 2000). To test whether inhibition of phosphorylation affects the distribution and dynamics of HMGA1a-GFP proteins, we blocked PKC activity by a cell permeable PKC-specific inhibitor and p³⁴ activity by roscovitine. Inhibition of both protein kinases resulted in a more diffuse distribution of HMGA1a-GFP (Fig. 7A,a,b). FRAP analyses of heterochromatin after drug treatments showed that inhibition of PKC led to a significantly increased mobility of HMGA1a-GFP compared with untreated cells (Fig. 7B, Table 1). After inhibition of PKC, full recovery of a bleached region already occurred in 17-25 seconds. A similar increased mobility was measured after p³⁴ inhibition by roscovitine (Fig. 7B, Table 1). Here, the fluorescence recovery of the bleached region occurred after 14-23 seconds. These experiments demonstrate that reduced phosphorylation by PKC and/or p³⁴ weakens the interaction of HMGA1a with DNA/chromatin in vivo.

To test whether acetylation has an impact on HMGA1a-GFP dynamics, histone deacetylases (HDACs) were inhibited by trichostatin A (TSA). This treatment also resulted in a less-confined distribution of HMGA1a-GFP (Fig. 7A,c). In FRAP-experiments of TSA-treated cells, fluorescence recovery of a bleached region was complete in 20-28 seconds compared with approximately 37 seconds in untreated cells (Fig. 7B, Table 1). These results indicate that chromatin acetylation increases HMGA1a dynamics in vivo.

Thus, after kinase or HDAC inhibition, the kinetic differences of HMGA1a-GFP found in heterochromatin were lost. This suggests that the protein kinases and histone deacetylases are involved in targeting HMGA1a to heterochromatin.

To rule out non-specific or indirect effects by drug-induced kinase inhibition, we used HMGA1a-GFP fusion proteins that were point mutated at the phosphorylation sites T21, T53 and T78 by substituting the threonine by alanine residues (Siino et al., 1995). At T53 and T78, HMGA1a proteins are known to be heavily phosphorylated, especially during mitosis (Nissen et al., 1991). Mutants T21A, T53A and T78A bore a point mutation in one phosphorylation site, T21,78A, T21,53A and T53,78A were double-point mutants and all three phosphorylation sites were mutated in the triple mutant T3×A (T21,53,78A). Mobility and distribution of these point-mutants were analyzed by FRAP during interphase and mitosis.

The cellular distribution of the mutant HMGA1a-GFP proteins is depicted in Fig. 8. Compared with the wild-type HMGA1a-GFP the single, double or triple mutants showed a less-prominent accumulation in heterochromatin, and during mitosis the mutated fusion proteins were enriched in the cytoplasm.

The mobility of the point-mutated proteins in heterochromatin (located within the nucleoplasm) was investigated. FRAP analyses of the single mutants T21A and T78A revealed a slightly, but significantly, increased mobility (Fig. 9, Table 1). By contrast, the kinetic behavior of the T53A point mutant was identical to that of wild-type HMGA1a-GFP (Fig. 9, Table 1). These results indicate that both phosphorylation sites T21 and T78, but not T53, are involved in regulating binding to heterochromatin. When we analyzed the double-point mutants (T21,53A; T21,78A; T53,78A) by FRAP, the T21,53A and the T21,78A mutants showed a significant increase in mobility compared with the wild-type protein (Fig. 9, Table 1). This indicates that T21 acts cooperatively with other sites in regulating heterochromatin binding. By contrast, the recovery kinetics of the double-point-mutant T53,78A was comparable with that of the single point mutant T21A or T78A (Fig. 9, Table 1), respectively. Furthermore, additional mutation of T53 reduced the cooperative effect of the T21,78A double mutant in the triple mutant T3×A. Significantly, in cells expressing T3xA, fluorescence recovery was slower than that of T21,78A and identical to that of the wild-type protein (Fig. 9, Table 1). These results indicate that phosphorylation of T53 does not directly contribute to HMGA1a heterochromatin binding in interphase cells but weakens the effect of phosphorylations at T21 and T78. However, they also suggest that different combinations of phosphorylation patterns act cooperatively (in the case of T21,53A and T21,78A) in stabilizing the heterochromatin binding but also non-cooperatively (in T3×A) in weakening heterochromatin binding.

In mitotic cells, all GFP-fusion proteins mutated at phosphorylation sites were preferentially located to chromosomes, but were also found in the cytoplasm, indicating a reduced binding to chromosomes (Fig. 8). FRAP-analyses of chromosomes showed that all point-mutated fusion proteins have significantly altered dynamics during mitosis (Fig. 9, Table 2). Compared with the wild-type HMGA1a-GFP, fluorescence recovery of a bleached chromosome occurred in a quarter of the time taken in T21A (Table 2). For T53A and T78A it occurred in a third of the time (Table 2). Comparable with the kinetic properties found in interphase heterochromatin, the double-mutated T21,53A and T21,78A fusion proteins displayed cooperative effects resulting in a faster recovery of bleached chromosomes compared with the single mutants. Likewise, the mobility of the T53,78A mutant or the triple mutated T3×A was reduced compared with the T21,78A double mutant (Table 2). Thus, as in interphase cells, this indicates the same cooperative and non-cooperative influence of different phosphorylation patterns. However, by contrast with interphase cells, our results show that all threonines next to each AT hook play a major role in regulating HMGA1a DNA/chromatin binding during mitosis.

Recent studies using photobleaching techniques have shown that many chromatin-associated proteins, such as histone H1 (Lever et al., 2000; Misteli et al., 2000), HMGN proteins (Catez et al., 2002; Phair and Misteli, 2000), HMGB proteins (Bianchi and Beltrame, 2000), topoisomerase II (Christensen et al., 2002) or heterochromatin protein 1 (Cheutin et al., 2003), move rapidly through the nucleus of a living cell. Here we present data on the dynamic properties of another class of HMG proteins, the HMGA1a proteins. By transient expression of GFP-tagged wild type and mutant fusion proteins in human cells, we have studied the in vivo distribution and dynamic interaction with DNA/chromatin by using FRAP experiments.

HMGA1a-GFP fusion proteins were distributed throughout the nuclear interior, excluding the nucleoli, with a distinct enrichment in heterochromatin blocks. Double-localization experiments localizing histone H1 or nascent transcripts revealed that HMGA1a-GFP proteins were preferentially enriched in transcriptionally inactive, condensed chromatin. Likewise, HMGA1a-GFP proteins were associated with chromosomes during mitosis. This distribution corresponds to previous immunfluorescence studies (Amirand et al., 1998; Martelli et al., 1998) and earlier observations, describing that HMGA1a proteins bind preferentially to A-tracts within heterochromatic regions (Radic et al., 1992; Reeves and Elton, 1987; Strick and Laemmli, 1995; Zhao et al., 1993). Nevertheless, we found that HMGA1a-GFP fusion proteins associate dynamically with DNA/chromatin throughout the cell cycle. This implies that HMGA1a proteins do not represent an immobile constituent of either the nuclear matrix or of a chromosome scaffold. Comparable with this conclusion is the recently published finding that topoisomerase II, another member of the matrix or scaffold fraction, is highly dynamic (Christensen et al., 2002). However, as for topoisomerase II, the mobility of HMGA1a does not rule out a dynamic interaction with such structures.

Conversely, the HMGA1a mobility provides us with scope to explain the multiple functions of these proteins in regulating transcription and as architectural elements of the chromatin. Despite their preferential association with heterochromatin, the mobile HMGA1a proteins can move to transcription sites and participate in the transcriptional regulation of several genes. In addition, the HMGA1a mobility would also ensure a dynamic interplay with other chromatin components, as discussed elsewhere (Käs et al., 1993; Zhao et al., 1993). For example, HMGA proteins were considered as anti-repressor molecules that could locally displace inhibitory chromatin proteins, such as histone H1, at AT-rich SAR sequences.

Our investigations using point-mutated fusion proteins show that distribution and dynamics are essentially governed by functional AT-hook motifs. Actually, all three AT hooks contribute to DNA/chromatin-binding in vivo. AT hooks I and II are the main mediators of DNA binding, whereas AT hook III plays only a secondary, but nevertheless cooperative part. For proper binding to chromatin in vivo, two functional AT hooks are necessary and sufficient. This conclusion is in accordance with several prior in vitro investigations (Claus et al., 1994; Frank et al., 1998; Maher and Nathans, 1996; Yie et al., 1997).

As the HMGA proteins are one of the most extensively modified nuclear proteins (Reeves, 2001), we also investigated the contribution of modifications on the HMGA1a DNA/chromatin binding. Treatment of cells with drugs inhibiting kinases, specifically protein kinase C (PKC) and p³⁴-kinase, or histone deacetylases (HDACs), increased the HMGA1a-GFP mobility in heterochromatin. Interestingly, the different kinetic properties found in nucleoplasmic or perinucleolar heterochromatin regions of untreated cells were lost after drug treatment. This observation indicates that phosphorylation and acetylation of HMGA1a modulates its binding to heterochromatin. In fact, previous studies identified PKC as a possible kinase acting on HMGA proteins in vitro (Schwanbeck and Wisniewski, 1997) and in vivo (Banks et al., 2000; Xiao et al., 2000), and the increased dynamics after inhibition of HDACs is consistent with the finding that HMGA1a acetylation results in destabilization of the enhanceosome (Munshi et al., 1998).

Our data suggest that the consequence of HMGA1a hypophosphorylation is a reduced affinity to heterochromatin. Conversely, we can assume that hyperphosphorylation should increase HMGA1a binding to heterochromatin. This assumption is supported by a decreased mobility of HMGA1a-GFP over mitotic chromosomes, when HMGA1a proteins are known to be hyperphosphorylated (Nissen et al., 1991).

FRAP studies using point-mutated proteins, where one or several mitosis-specific phosphorylation sites were replaced, confirmed the regulatory role of HMGA1a phosphorylation on its dynamic behavior. All point-mutated fusion proteins studied showed increased mobility during mitosis, which strongly indicates that reduced phosphorylation (mimicked here by the mutations T21A, T53A, T78A, or double and triple mutations at these sites) reduces binding to DNA/chromatin. Notably, this implies that besides the known mitosis-specific phosphorylation sites T53 and T78 (Nissen et al., 1991), the threonine at position 21 plays a crucial role for the regulation of HMGA1a binding to chromosomes. Thus, all threonines that are next to the AT-hook-binding motifs are involved in regulating the chromatin binding during mitosis and phosphorylation of these threonines increases binding of HMGA1a to chromosomes in vivo.

FRAP experiments also showed that some of these threonines play a crucial role for heterochromatin binding during interphase. The threonines at the positions 21 and 78 appear to be most relevant to this phenomenon, which have already been shown to be phosphorylated in vitro and in vivo (Banks et al., 2000; Schwanbeck and Wisniewski, 1997; Xiao et al., 2000). Alone, mutation of the threonine at position 53 next to AT hook II does not play a significant role for heterochromatin binding. However, investigation of double- or triple-point-mutated proteins by FRAP indicates that phosphorylation at T53 regulates heterochromatin binding in combination with T21 and T78. These FRAP data show that different combinations of phosphorylations result in different, but not necessarily cooperative, changes in HMGA1a dynamics. Consequently, different combinations of phosphorylations lead to different kinetic properties and fine-tune HMGA1a dynamics especially in heterochromatin. In sum, these data imply that, compared with euchromatin, the increased residence times found in heterochromatin and in mitotic chromosomes are caused by an increased phosphorylation of HMGA1a proteins. However, whether HMGA1a hyperphosphorylation induces chromatin condensation in vivo still has to be examined. Similar to HMGA1a, heterochromatin binding of histone H1 is also affected by the phosphorylation status (Contreras et al., 2003; Dou et al., 2002). However, in contrast with HMGA1a proteins, histone H1 in heterochromatin is in an unphosphorylated state (Contreras et al., 2003).

Many previous investigations showed that phosphorylation of HMGA proteins reduces DNA binding in vitro (reviewed in Reeves and Beckerbauer, 2001). In vivo, DNA is packed into chromatin and HMGA proteins were also shown to interact with nucleosomes (Reeves et al., 2000). Therefore, we can assume that phosphorylated HMGA1a proteins preferentially interact with positively charged histones, while dephosphorylated HMGA1a proteins preferentially interact with negatively charged DNA. This phosphorylation-dependent movement between DNA binding and histone binding might contribute to activation and inactivation of chromatin sites. In such a model, hyperphosphorylated proteins should bind with higher affinity to histones and be specific for condensed chromatin, whereas hypophosphorylated HMGA1a proteins preferentially bind to DNA and should be enriched in decondensed chromatin. Based on the results about phosphorylation patterns found in HMGA1 proteins of Chironomus, such a model has already been discussed (Schwanbeck and Wisniewski, 1997). Our results strongly support this model: (1) HMGA1a proteins are highly mobile in vivo; (2) the AT hook ensures a dynamic binding to DNA; (3) this mobility is then fine tuned by different modification patterns (i.e. different combinations of phosphorylation). Consequently, different dynamic populations might then compete with other proteins for DNA- or histone-binding sites in either euchromatin or heterochromatin. A similar competition model has recently been proposed for heterochromatin protein 1 (HP1), which is also dynamically associated with heterochromatin (Cheutin et al., 2003). The dynamic behavior of HMGA1a in heterochromatin strongly supports the model that heterochromatin is a dynamic structure and accessible to regulatory proteins (Cheutin et al., 2003).

This work received support by a graduate program from the Deutsche Forschungsgemeinschaft (GK639). We thank Ray Reeves for kindly providing HMGA1 antibodies.