Human ING1 Proteins Differentially Regulate Histone Acetylation

ING1 proteins are nuclear, growth inhibitory, and regulate apoptosis in different experimental systems. Here we show that similar to their yeast homologs, human ING1 proteins interact with proteins associated with histone acetyltransferase (HAT) activity, such as TRRAP, PCAF, CBP, and p300. Human ING1 immunocomplexes contain HAT activity, and overexpression of p33(ING1b), but not of p47(ING1a), induces hyperacetylation of histones H3 and H4, in vitro and in vivo at the single cell level. p47(ING1a) inhibits histone acetylation in vitro and in vivo and binds the histone deacetylase HDAC1. Finally, we present evidence indicating that p33(ING1b) affects the degree of physical association between proliferating cell nuclear antigen (PCNA) and p300, an association that has been proposed to link DNA repair to chromatin remodeling. Together with the finding that human ING1 proteins bind PCNA in a DNA damage-dependent manner, these data suggest that ING1 proteins provide a direct linkage between DNA repair, apoptosis, and chromatin remodeling via multiple HAT.ING1.PCNA protein complexes.


3
The ING1 candidate tumor suppressor gene expresses a family of alternatively spliced mRNAs encoding proteins that localize to the nucleus and that are growth inhibitory (1)(2)(3)(4). The locus of ING1 maps to chromosome 13q33-34 (5), a site frequently associated with loss of heterozygosity in several types of cancers (6,7). In human cells, p47 ING1a and p33 ING1b are the major ING1 splicing isoforms expressed (8) and p33 ING1b is the most intensively characterized isoform to date. Suppression of p33 ING1b expression promotes focus formation and growth in vitro, and tumor formation in vivo, while ectopic overexpression of this protein was shown to block cell cycle progression by arresting transfected cells at G1 of the cell cycle (2,4). Clinical data have shown that reduced levels of p33 ING1b are seen in primary breast tumors (9), lymphoid malignancies (10), testis (11) and squamous cell cancers (12,13), consistent with ING1 acting as a class 2 tumor suppressor (14). p33 ING1b also displays properties of a regulator of apoptosis in different experimental systems (15)(16)(17)(18)(19)(20). Both the apoptotic and cell cycle regulatory properties of p33 ING1b may involve the tumor suppressor p53, with which p33 ING1b and the closely related p33 ING2 were found to be capable of physically and/or functionally interacting (16,(20)(21)(22).
Finally, we have recently reported that p33 ING1b was able to bind to PCNA in a DNA damageinducible manner that was directly linked to the ability of p33 ING1b to induce apoptosis (17).
Recent studies suggest that human ING1 proteins might be involved in chromatin remodelling functions via physical association with both histone acetyltransferases (HATs) and histone deacetylases (HDACs). We initially reported that an ING1 yeast homolog protein, Yng-2, was able to interact with Tra1 (23), a protein that is part of HAT complexes such as SAGA and NuA4 (24,25). Recently, while Yng-2 was shown to be essential for a Nu4A-mediated HAT activity controlling cell proliferation (26,27), the human p33 ING1b was found to be functionally and physically linked to HDAC1 (28,29). Despite these observations, neither the biochemical by guest on March 24, 2020 http://www.jbc.org/ Downloaded from 4 role(s) of human ING1 proteins in HAT-related functions nor the biological significance of these functions is fully understood. Furthermore, there is no information published regarding the functions and biological properties of p47 ING1a , the other major human isoform of ING1. We have recently reported 1 that different ING1 proteins displayed isoform-dependent apoptotic properties correlated with differential binding affinity to chromatin. These observations led us to ask in the current study: A) whether human ING1 proteins could physically associate with HATs in the same way as yeast Yng-2 associates with Tra1 (23); B) whether human ING1 immunocomplexes were able to co-precipitate HAT activity; C) whether different human ING1 isoforms would display differential HAT/HDAC properties correlated with their differential binding to HATs, and if so D) whether these ING1 biochemical functions correlated with any apoptotic effect. Finally, since p33 ING1b was reported to be involved in UV-induced damage responses (30) and we observed an UV-dependent physical interaction between ING1 proteins and PCNA (17), which was functionally relevant for UV-mediated apoptosis (17), we asked E) whether ING1 proteins could alter the recently noted association between p300 and PCNA (31), an association that suggests a linkage exists between the repair of UV-damaged DNA and the global regulation of gene expression through histone acetylation (31).
To address these questions, we have characterized the molecular, biochemical and biological properties of the two major human ING1 isoforms, p47 ING1a and p33 ING1b , regarding their roles in functions related to the acetylation of chromatin. We present evidence indicating that similar to yeast ING1 immunocomplexes (23), human ING1 immunocomplexes coprecipitate HAT activity. We show that ING1 proteins physically interact with proteins present in complexes containing HAT activity such as CBP, PCAF, p300 and TRRAP. TRRAP, the 5 human homolog of yeast Tra1 which interacts with Yng-2 (23), is a 434 kDa c-Myc-interacting nuclear protein that binds the c-Myc amino terminus and the E2F-1 transactivation domain, and is an essential cofactor for c-Myc and adenovirus E1A-mediated oncogenic transformation (32); TRRAP recruits at least one HAT, hGCN5, to a complex containing c-Myc (24,33).
These physical associations, HAT activities and biological properties associated with ING1 proteins were found, in the present study, to be isoform-dependent. Overexpression of p33 ING1b , but not p47 ING1a , induced apoptosis and hyperacetylation of histones H3 and H4 both in vitro and in vivo at a single cell level. Conversely, ectopic up-regulation of p47 ING1a resulted in a decrease of the levels of histone acetylation in vitro and in vivo accompanied by no changes in the percentage of apoptosis compared to the empty vector control. In agreement with this observation, we have found that p47 ING1a avidly binds to the histone deacetylase HDAC1.
Finally, we present evidence indicating that p33 ING1b affects the degree of physical association between PCNA and p300, an association that has been proposed to link DNA repair to chromatin remodelling functions (31).

EXPERIMENTAL PROCEDURES.
Cell culture. Primary normal human diploid fibroblasts (Hs68; ATCC CRL#1635) and were rinsed thrice with PBS and exposed to UV (25 J/m 2 ) as described previously (17,18). (8) were subcloned into pCI vector (Promega). All 6 experiments included parallel controls of cells transfected with green fluorescent protein (GFP) (Clontech) expression plasmids to determine the proportion of transfected cells. Cells were electroporated as described previously (17,18). For western blots ( Figure 6), cells were cotransfected with ING1 and GFP expression constructs at an ING1:GFP plasmid ratio of 4:1.

Plasmids and transfections. ING1 cDNAs
Antibodies, Immunoprecipitation and Western blot assays. ING1 polyclonal rabbit antibodies (2,34), the four mouse CAb ING1 monoclonals (35), and the c-Fos mouse monoclonal antibodies (36), have been characterized previously. We also used rabbit polyclonal Microscopy and microinjection of somatic cells. Cells were cultured, microinjected, fixed and mounted as described (17,37). Acetylated histones were visualized using the above-mentioned antibodies according to the supplier's recommendations. After washing, cells were incubated with the secondary antibodies goat anti-rabbit IgG (Cy3, Chemicon) or goat anti-mouse IgG (Alexa 488, Cedarlane or Cy5, Chemicon). After rinsing, the samples were mounted in 1 µg/ml paraphenylenediamine in PBS/90% glycerol that also contained the DNA specific dye DAPI at 1 µg/ml. Imaging was performed using a 14-bit cooled CCD camera (Princeton Instruments) mounted on a Leica DMRE immunofluorescence microscope. For signal density quantitation, the nuclear signal of acetylated histones was integrated for injected and non-injected cells, using ERGOvista v4.4 software.

FACS analysis and cell sorting. For analysis of apoptosis, cells electroporated with ING1
expression constructs were harvested at 48 hours, fixed in 70% ethanol/PBS, on ice for 1 hour after which they were subjected to analysis or were kept at -20°C for no more than one week.
Before analysis using a Becton Dickinson FACS-scanner, ethanol was removed and cells were resuspended in PBS for 10 minutes, after which they were pelleted, the PBS was removed and the cells were treated with staining solution (5µg/mL of propidium iodide (Sigma), 1mg/mL of  (23), we asked whether human ING1 proteins would be able to bind to TRRAP, the human homolog of Tra1, as well as to other proteins of the GNAT and MYST HAT superfamilies (24,32). To test this hypothesis, we made use of immunoprecipitation-western (IP-W) assays. As shown in Fig. 1A-D, immunoprecipitation of endogenous ING1 proteins from lysates of SNB19 cells co-precipitated several HATs, including TRRAP, PCAF and CBP. Although we were able to find a weak physical association between ING1 proteins and PCAF, which is a member of the hGNC5 family of HATs (24), we were unable to find any interaction between ING1 proteins and hGCN5 (Fig. 1C). In order to confirm that the interaction between ING1 and one of the HAT complex proteins was specific, we tested whether anti-CBP, but not rabbit preimmune control immunoprecipitates, was able to coimmunoprecipitate ING1 proteins. As shown in Fig. 1E, a weak but reproducible band in the anti-CBP lane corroborated the idea that the ING1:CBP interaction was specific. Similar results were obtained in primary fibroblast cells (data not shown). Because of this apparently weak association between CBP and ING1 proteins, we wanted to test whether this interaction was increased upon ectopic up-regulation of these proteins. For this purpose, we performed IP-W assays on lysates of primary human fibroblasts overexpressing these proteins upon cotransfection of constructs encoding CBP and either p47 ING1a , p33 ING1b or empty expression vector (indicated as a, b and v in Fig. 2). IP-W assays showed that while both ING1 isoforms bound CBP, p33 ING1b appeared to bind it much more avidly than p47 ING1a ( Fig. 2A). To confirm that the difference in the affinity of ING1 isoforms by CBP was not due to different expression levels of these proteins (i.e., as consequence of differences in the efficiencies of transfection), we  (17), CBP is closely related to p300 (38)(39)(40), and a recent report has indicated that p300 was able to bind PCNA through an unidentified nuclear protein (31) Figure 3C shows that ING1 transfection resulted in the expected overexpression of p47 ING1a and p33 ING1b in the absence and in the presence of UV. Fig. 3D shows that PCNA levels were not altered appreciably under these conditions. Aliquots of the same cell lysates blotted in figures 3C and 3D were immunoprecipitated with anti-p300 antibodies and immunoprecipitates were blotted with anti-PCNA antibodies (Fig. 3B). As shown in Fig. 3B, p300 and PCNA were found together in complexes that were not altered by UV alone (lanes with vector transfected) as reported previously (31). However, while high levels of p47 ING1a did not appreciably alter the amount of PCNA complexed with p300 and low levels of p33 ING1b had a modest effect, overexpression of higher levels of p33 ING1b selectively interfered with the p300:PCNA interaction following UV exposure (compare lane b-to lane b+ in Fig. 3B).

ING1 immunoprecipitates contain HAT activity. Since human ING1 proteins interacted with
HATs ( Fig. 1-3), and yeast ING1 immunoprecipitates contained HAT activity (23) Figure   4C is a Coomassie gel of samples from Figure 4B. As shown in these figures, anti-ING1 antibodies (αI) precipitated abundant HAT activity, whereas preimmune sera (PI) did not. The autoradiograph shown in Fig. 4A was overexposed to demonstrate that even though Fos has been reported to bind to CBP (41), Fos antibodies precipitated less than 10% of the HAT activity seen in ING1 immunoprecipitates, consistent with ING1 isoforms associating with multiple HAT complexes. This HAT activity was ING1-specific since blocking the anti-ING1 antibodies by preincubation with a GST-ING1 fusion protein markedly decreased this co-precipitated HAT activity (Fig. 4A, compare αI with αI+P).

Different ING1 isoforms are in complexes displaying different histone acetylation
properties. Since ING1 isoforms differentially interacted with different HATs (Fig. 2 and 3 As shown in Fig. 4B, similar results were obtained by quantitation of HAT activity from total lysates of ING1 transfectants. These data are consistent with data of Fig. 1 and support the idea that ING1 proteins contribute to the regulation of multiple HAT complexes. The magnitudes of these differences as estimated by scanning densitometry were, on average, a 60% increase and a 45% decrease for ING1b and ING1a transfected cells, respectively, but these relatively modest changes are most likely due to low transfection efficiencies of human fibroblasts that typically range from 10-40%. Alternatively, the inhibitor may serve to abolish the proposed HDAC-associated properties of p33 ING1b (28,29). Table 1  We are currently defining regions of the ING1 proteins that are responsible for such protein:protein interactions.

DISCUSSION.
In this study we demonstrated that similar to yeast ING1 proteins (23), human ING1 proteins associate with HATs ( Fig. 1-3), co-precipitate HAT activity (Fig. 4) and regulate the acetylation of histones in vitro and in vivo (Fig. 4-6). These properties were shown to be ING1 isoform-dependent ( Fig. 4-6) and correlated with differential effects of ING1 isoforms on apoptosis (Fig. 7). Immunoprecipitation studies allowed us to identify different HAT complexes such as those containing TRRAP, CBP, p300 and PCAF as ING1-interacting complexes (Figs. 1-3) and overexpression studies supported the idea that ING1 isoforms interact directly with the HAT proteins themselves. However, this point must be confirmed by alternative approaches and is currently being addressed using purified proteins. The interaction between p300 and ING1 led us to find that ING1 proteins associated in a complex with both PCNA and p300 (Fig. 3). This association was affected by the levels ING1 in isoform-dose-and UV-dependent manners (Fig.   3). By means of several independent studies, we identified histones H3 and H4 as targets of the HAT regulatory properties displayed by ING1 isoforms (Fig. 4 and 5). While p33 ING1b increased the level of acetylation of these histones, p47 ING1a exerted the opposite effect (Fig. 4-6). Finally, we observed that these ING1-mediated HAT regulatory functions directly correlated with ING1 isoform-dependent apoptotic effects (Fig. 7).
The correlations between the differential HAT and apoptotic properties displayed by  (21), as well as survival and apoptotic genes such as Bax (22) and others (4,22). At present it is unclear whether the effect of p47 ING1a (Table 1) is through competitive inhibition of the function of p33 ING1b in complexes containing HAT activity, through the activation of HDAC activity, or both. Consistent with the former possibility, p47 ING1a is clearly capable of interacting with CBP, although with considerably less avidity than p33 ING1b (Fig. 2). Also, HDAC1 has been observed to interact in a complex with the p33 ING1b isoform (28,29) and we have noted that HDAC1 interacts avidly with p47 ING1a and to a slightly lesser extent, with p33 ING1b (Fig. 3A).
The presence of ING1 proteins in complexes containing p300 and PCNA (Fig. 3B) provides additional support for a recent report linking chromatin remodelling to DNA repair through an interaction between PCNA and p300 (31); an association that was proposed to occur through an unidentified nuclear protein/s (31). If p33 ING1b is the protein linking PCNA with p300, overexpression of p33 ING1b would be expected to saturate binding sites on both proteins, by guest on March 24, 2020 http://www.jbc.org/ Downloaded from inhibiting the formation of complexes containing all 3 proteins. Such p33 ING1b -induced dissociation was seen clearly at higher levels of p33 ING1b expression (Fig.3). Since human ( Fig.   1-3) and yeast ING1 proteins (23)(24)(25) bind to proteins present in different HAT complexes, and human ING1 proteins also seem to be involved in UV-induced cell damage responses (17,30), the ING1 family of proteins may constitute an important link between chromatin remodelling and DNA repair. A report linking the TIP60 histone acetylase directly to DNA repair and apoptosis (43) strengthens this idea further. TIP60, a member of the MYST family of HATs, is related to the yeast Esa1 (24), which interacts with a yeast homolog of the mammalian p33 ING1b (23). Esa1 preferentially acetylates histone H4 (24), which is also acetylated by anti-ING1 immunoprecipitates (Fig. 4).
The association observed between ING1 and CBP ( Fig. 1 and 2) also helps to clarify why p33 ING1b , although a potent growth inhibitor when overexpressed in normal cells (2), is unable to block the growth of cells expressing SV40 large T antigen (1). Since CBP is an obligate cellular target of the large T antigen oncoprotein (40)    Note that although both ING1 proteins bind to CBP, p33 ING1b binds considerably more avidly.