A Requirement for Dimerization of HP1Hsα in Suppression of Breast Cancer Invasion*

The development and progression of cancer is controlled by gene expression, often regulated through chromatin packaging. Heterochromatin protein 1Hsα (HP1Hsα), one of three human HP1 family members, participates in heterochromatin formation and gene regulation. HP1Hsα possesses an amino-terminal chromodomain, which binds methylated lysine 9 of histone H3 (meK9 H3), and a carboxyl-terminal chromoshadow domain (CSD) that is required for dimerization and interaction with partner proteins. HP1Hsα is down-regulated in invasive metastatic breast cancer cells compared with poorly invasive nonmetastatic breast cancer cells. Expression of EGFP-HP1Hsα in highly invasive MDA-MB-231 cells causes a reduction in in vitro invasion, without affecting cell growth. Conversely, knock-down of HP1Hsα levels in the poorly invasive breast cancer cell line MCF-7 increased invasion, without affecting cell growth. To determine whether functions of the CSD were required for the regulation of invasion, mutant forms of HP1Hsα were expressed in MDA-MB-231 cells. A W174A mutation that disrupts interactions between HP1Hsα and PXVXL-containing partner proteins reduced invasion similar to that of the wild type protein. In contrast, an I165E mutation that disrupts dimerization of HP1Hsα did not decrease invasion. No gross changes in localization and abundance of HP1Hsβ, HP1Hsγ, and meK9 H3 were observed upon expression of wild type and mutant forms of HP1Hsα in MDA-MB-231 cells. Taken together, these data demonstrate that modulation of HP1Hsα alters the invasive potential of breast cancer cells through mechanisms requiring HP1 dimerization, but not interactions with PXVXL-containing proteins.

Mortality from breast cancer occurs by the spread of cancer cells to secondary sites within the body, a process called metastasis (1). Breast cancer cells must acquire several properties in order to disseminate from the primary tumor, including the ability to degrade and migrate through the extracellular matrix, a process called invasion (2). Invasion is one of the first steps in the metastatic cascade and is a strong indicator of tumor progression. Two classes of proteins have been identified that regulate metastasis progression: activators and suppressors (3). Metastasis activators are genes that promote metastasis, whereas metastasis suppressors inhibit metastasis. In contrast to tumor suppressors, metas-tasis suppressors do not affect growth of the primary tumor (3). Although several genes have been identified that regulate metastasis, clinically valuable predictive or prognostic molecular markers for metastasis have yet to be determined (4 -6). Currently, the best indicator for metastasis is lymph node micrometastases (4). However, microarray-based studies to identify transcription profiles that predict metastasis are in development (5,7).
In an attempt to identify genes with altered expression in breast cancer metastasis, a differential display analysis was performed comparing gene expression between poorly invasive nonmetastatic and highly invasive metastatic breast cancer cells (8). The mRNA encoding heterochromatin protein 1 Hs␣ (HP1 Hs␣ ) 3 was found to be down-regulated (1.5fold) in highly invasive metastatic breast cancer cell lines compared with poorly invasive nonmetastatic breast cancer cell lines (8,9). Consistent with this reduction in mRNA, HP1 Hs␣ protein levels showed an even larger disparity (7.1-fold) between the two types of breast cancer cells (9). HP1 Hs␣ protein was also less abundant in metastatic tissues than in primary breast cancer tissues, suggesting that the down-regulation in highly invasive metastatic breast cancer cell lines recapitulate trends that were observed in clinical samples (9). HP1 is a highly conserved protein originally identified in Drosophila melanogaster as a component of chromatin near centromeres (10,11). HP1 proteins are classically known to silence genes that have been juxtaposed to centric chromatin, and the degree of silencing is dependent on HP1 dosage (12,13). Three HP1 family members have been identified in humans: HP1 Hs␣ , HP1 Hs␤ , and HP1 Hs␥ . All HP1 family members consist of two conserved domains, the chromodomain (CD) and the chromoshadow domain (CSD), that are separated by a less conserved hinge region. The CD, within the amino-terminal region, folds into a structure containing three ␤-sheets and an ␣-helix that form a hydrophobic protein interaction pocket (14,15). The HP1 CD has been shown to specifically interact with di-and trimethylated lysine 9 of histone H3 (meK9 H3), an epigenetic mark generated by SET (suppressor of variegation, enhancer of zeste and Trithorax) domain-containing histone methyltransferases (15)(16)(17)(18). This interaction targets HP1 proteins to specific regions of the genome that are enriched in meK9 H3, such as those near centromeres (16,19).
The CSD, at the carboxyl-terminal region, has an amino acid sequence and structure similar to that of the CD (20 -22). A major distinction between the CD and CSD is that the CSD forms homo-and heterodimers with other HP1 proteins through an ␣-helix (20,21,23). * This work was supported by Grant DAMD17-02-1-0424 from the Dept. of Defense The dimerization of CSDs forms an interaction platform for proteins containing the amino acid sequence motif PXVXL (21,22). Many different types of nuclear proteins contain PXVXL motifs, including nuclear architecture proteins such as the lamin B receptor, transcriptional regulators such as KRAB-associated protein 1 (KAP1), and chromatin assembly and modifying proteins such as chromatin assembly factor 1 (CAF1p150) (24). However, there have been proteins identified that bind to HP1 but do not utilize the PXVXL platform interaction, such as BRM-related gene 1 (BRG1) and suppressor of variegation 3-9 homolog 1 (SUV39h1) (25,26).
In this report, the function of HP1 Hs␣ has been investigated by assaying invasion and growth of breast cancer cells upon modulation of HP1 Hs␣ levels. Structure/function analysis has revealed that HP1 Hs␣ dimerization is essential for regulating invasion. Data presented here are consistent with the hypothesis that HP1 Hs␣ functions as a breast cancer metastasis suppressor.
Adenoviral Constructs-The I165E and W174A mutants of HP1 Hs␣ were generated by site-directed mutagenesis (QuikChange, Invitrogen) using the pEGFP-HP1 Hs␣ as a template (9). The primers used to generate the I165E mutation are: 5Ј-CCACAAATTGTGGAAGCATTT-TATG-3Ј and 5Ј-CATAAAATGCTTCCACAATTTGTGG-3Ј. The primers used to generate the W174A mutation are: 5Ј-GAGACTGA-CAGCGCATGCATATCC-3Ј and 5Ј-GGATATGCATGCGCTGT-CAGTCTC-3Ј. The mutant and wild-type EGFP-HP1 Hs␣ were cloned into the Ad5 RSV K-NpA shuttle vector (University of Iowa Gene Transfer Vector Core). As a control, the enhanced green fluorescent protein (EGFP) gene was fused to a nuclear localization signal (isolated from pShooter pCMV/myc/nuc plasmid, Invitrogen) and cloned into the Ad5 RSV K-NpA shuttle vector. The adenoviral constructs were packaged into virus and used for infection of breast cancer cells (University of Iowa Gene Transfer Vector Core).
A short hairpin RNA interference (RNAi) construct for knock-down of HP1 Hs␣ (shHP1 Hs␣ ) was designed using the "Shagging PCR Protocol" and the "RNAi oligo retriever" (27,28). The resulting sequences were generated as reverse PCR primers (reverse primer 5Ј-AAAAAAAATATTCCACT-TATCCCTCAACCACGCACCAAGCTTCGCGCGTGGTTAAGGGA-CAAGTGGAATATCGGTGTTTCGTCCTTTCCACAA-3Ј) and used in a PCR reaction in combination with a forward U6 primer and a pGEM-U6 promoter template (kind gift from Greg Hannon, Cold Spring Harbor Laboratory). The PCR product was cloned into the pGem-T vector (Invitrogen) and subsequently cloned into the Ad5 RSV K-NpA shuttle vector. shHP1 Hs␣ was designed to target the HP1 Hs␣ mRNA starting at position 272. The shRNA construct was packaged in combination with a viral backbone containing EGFP in the E3 region driven by a RSV promoter for identification of infected cells (University of Iowa Gene Transfer Vector Core). Negative controls consisted of a short hairpin RNAi construct against GFP (shGFP) expressed from either a U6 or CMV promoter (University of Iowa Gene Transfer Vector Core).
Adenoviral Infection-MDA-MB-231 or MCF-7 cells were grown to 50 -70% confluency in complete medium. Prior to infection, the medium was replaced with RPMI 1640 plus 1% fetal bovine serum and 10 mM gentamycin. Adenovirus was added, and the cells were incubated at 37°C with 5% CO 2 . After 2 h of incubation, complete medium, was added to the cells. Twenty-four hours post-infection, the medium was replaced with fresh complete medium. A multiplicity of infection (m.o.i.) of 60 plaque-forming units/cell was used for the EGFP-HP1 Hs␣ adenovirus to infect MDA-MB-231 cells. To ensure equivalent amounts of protein expression, the m.o.i. of the adenoviruses expressing EGFP-NLS, I165E, and W174A was adjusted so that the proteins were expressed in equal amounts to EGFP-HP1 Hs␣ . Therefore, the m.o.i. for the EGFP-NLS, I165E, and W174A adenoviruses was 5, 10, and 30 plaque-forming units/cell, respectively. The m.o.i. used for infection of MCF-7 cells with both the shGFP and shHP1 Hs␣ adenoviruses was 62 plaque-forming units/cell.
In Vitro Invasion Assays-In vitro invasion of breast cancer cells was performed using the Membrane Invasion Culture System (MICS) as described previously (9,29). Relative percent invasion was calculated by taking the mean of the technical replicates within each MICS chamber, dividing the total number of invading cells by the total number of cells seeded times 100, and normalizing to the control condition. The experiments were repeated three or four times, and the p value was calculated using the Microsoft Excel t test function.
Growth Assays-Twenty-four hours post-infection, cells were collected and plated (5 ϫ 10 4 cells MDA-MB-231; 2 ϫ 10 5 MCF-7) in each well of a 6-well plate. Cells were collected for each sample from one well at specific time points, and the cell number was counted using a hemocytometer.
Cell Cycle Assays-Forty-eight hours post-infection, 1 ϫ 10 6 MDA-MB-231 cells were collected, resuspended in 1ϫ PBS, and fixed in 70% ethanol for 1 h at 4°C. Cells were then washed in 1ϫ PBS, treated with RNase A, stained with a final concentration of 50 g/ml propidium iodide, and incubated for 30 min at 4°C. The stained cells were analyzed by flow cytometer (BD Biosciences FACScan), and the numbers of cells in each phase of the cell cycle were analyzed.
BrdUrd Incorporation Assays-Forty-eight hours post-infection, 5 ϫ 10 5 cells were plated in 24-well plates on top of glass coverslips, incubated for 24 h (37°C with 5% CO 2 ), and then treated with 10 M BrdUrd (12, 6, 3, or 1 h). The cells were fixed with methanol, blocked with 1ϫ PBS, 1% bovine serum albumin, treated with 1.5 N HCl, and immunostained with ␣-BrdUrd antibody (Developmental Studies Hybridoma Bank, University of Iowa). Goat ␣-mouse rhodamine (Molecular Probes) was used as a secondary antibody. The cells were co-stained with 1 g/l DAPI (Sigma-Aldrich). Cells were visualized under a fluorescent microscope (Leica DMLB). The percent BrdUrd incorporated cells was calculated as the total number of BrdUrd positive cells/total number of cells (DAPI positive).

Expression of HP1 Hs␣ in MDA-MB-231
Cells-Because of the correlation between low levels of HP1 Hs␣ and breast cancer invasion/metastasis, our objective was to modulate the levels of HP1 Hs␣ in breast cancer cell lines and assay for effects on cell growth and invasion. Adenoviral constructs expressing either enhanced green fluorescent protein (EGFP) tagged with a nuclear localization signal (NLS) as a negative control or EGFP fused to HP1 Hs␣ were generated. To determine whether GFP altered the function of HP1 Hs␣ , the EGFP-HP1 Hs␣ fusion protein was expressed in Drosophila, for which functional assays exist (30). Three lines of evidence demonstrate that EGFP does not alter function of HP1 Hs␣ : 1) EGFP-HP1 Hs␣ localized to sites of endogenous HP1 on Drosophila salivary gland polytene chromosomes; 2) EGFP-HP1 Hs␣ enhanced silencing of genes near heterochromatin in a dosagedependent manner; and 3) EGFP-HP1 Hs␣ rescued lethality associated with mutations in the gene encoding HP1 (30).
To determine the effects of EGFP-HP1 Hs␣ expression in breast cancer cells, MDA-MB-231 cells were infected with adenovirus encoding EGFP-HP1 Hs␣ at levels similar to that of endogenous HP1 Hs␣ in MCF-7 cells, recapitulating physiological levels in noninvasive breast cancer cells (Fig. 1). As a negative control, EGFP-NLS was expressed in MDA-MB-231 cells at levels comparable with that of EGFP-HP1 Hs␣ ( Fig. 2A). The localization of EGFP-HP1 Hs␣ was examined in both fixed and unfixed cells. EGFP-HP1 Hs␣ localized to small discrete foci within the nucleus and co-localized with endogenous HP1 Hs␣ in fixed cells ( Fig. 3 and data not shown), the pattern anticipated from previous localization studies (9,19,31,32). In contrast, EGFP-NLS exhibited a diffuse nuclear localization pattern ( Fig. 3 and data not shown). Expression of EGFP-NLS did not alter endogenous HP1 Hs␣ expression as assayed by reverse transcriptase PCR, nor did EGFP-NLS cause a change in endogenous HP1 Hs␣ localization (data not shown). These data demonstrate that the EGFP-tagged proteins properly localize in the nucleus of MDA-MB-231 cells.
Previous experiments using MDA-MB-231 cells with stable expression of HP1 Hs␣ showed a reduction in invasion in vitro compared with controls (9). To determine how HP1 Hs␣ induces this change, we first needed to reproduce this effect using an adenoviral system that allowed for greater flexibility in protein expression. The effect of expressing HP1 Hs␣ by adenovirus infection on the invasive potential of MDA-MB-231 cells was measured using the membrane invasion culture system (29). MDA-MB-231 cells infected with a virus expressing EGFP-NLS did not have a significantly different invasive potential than uninfected cells (106 versus 100%, respectively; p ϭ 0.455) (Fig. 2B). MDA-MB-231 cells infected with EGFP-HP1 Hs␣ have a reduced invasive potential compared with uninfected cells (76 versus 100%, respectively; p ϭ 0.023) or compared with EGFP-NLS (72 versus 100%, respectively; p ϭ 0.015) (Fig. 2, B and C). These results provide evidence that HP1 Hs␣ is involved in the suppression of breast cancer cell invasion, consistent with previously published data (9).
Dimerization Is Required for HP1 Hs␣ Suppression of Invasion in MDA-MB-231 Cells-With the adenoviral system established, we investigated the mechanism of HP1 Hs␣ invasion modulation using HP1 Hs␣ with mutations in the CSD. The ability of HP1 proteins to dimerize is hypothesized to be important for heterochromatin formation and for recruitment of partner proteins. Thus, the CSD is a good candidate for analysis because amino acid substitutions exist that separate dimerization and partner proteins interactions (Fig. 2D) (20 -22). The isoleucine 165 (Ile-165) residue of HP1 Hs␣ is necessary for CSD dimerization. HP1 Hs␣ proteins with mutations in this residue (to an alanine, glutamate, or lysine) are subsequently unable to dimerize and bind PXVXL-containing proteins (20,22,33,34). Unlike the Ile-165 residue, substitution at the tryptophan 174 (Trp-174) residue disrupts interactions with PXVXL-containing proteins without disrupting HP1 Hs␣ CSD dimerization (20,22,33). The Trp-174 residue is required for stabilizing the interactions of the CSD with the proline and leucine residues of the PXVXL motif (22). When Trp-174 is mutated to an alanine or glutamate, dimerization of the HP1 Hs␣ CSD is retained, but the CSD is unable to interact with PXVXL-containing proteins (20,33).
Adenoviruses were constructed that expressed EGFP fused to either I165E or W174A. Viral infection was normalized so that equal amounts of EGFP-NLS, EGFP-HP1 Hs␣ , EGFP-I165E, and EGFP-W174A proteins were expressed in the MDA-MB-231 cells ( Fig. 2A). Patterns of the EGFP fusion proteins were examined to determine whether the mutations affected localization. Expression of the EGFP-I165E fusion protein resulted in a more diffuse nuclear localization pattern in the nucleus of both unfixed and fixed MDA-MB-231 cells compared with the localization pattern of EGFP-HP1 Hs␣ (Fig. 3 and data not shown). The EGFP-W174A fusion protein localized to discrete foci in the nucleus in both unfixed and fixed MDA-MB-231 cells, similar to the localization pattern observed with endogenous HP1 Hs␣ in MCF-7 cells (Fig. 3 and data not  shown). These staining patterns suggest that the localization of the I165E mutant protein is disrupted, whereas the localization of the W174A mutant protein is similar to that of wild-type HP1 Hs␣ .
To determine the ability of the HP1 Hs␣ mutant proteins to suppress invasion, an in vitro invasion assay was performed on cells expressing mutant forms of HP1 Hs␣ , with EGFP-NLS as a negative control (Fig. 2C). I165E was unable to suppress invasion of MDA-MB-231 cells compared with EGFP-NLS (100 versus 100%, respectively; p ϭ 1.0). Therefore, dimerization of the HP1 Hs␣ CSD is required for the ability of HP1 Hs␣ to suppress invasion in MDA-MB-231 cells. In contrast, W174A was able to suppress invasion of MDA-MB-231 cells compared with EGFP-NLS (68 versus 100%, respectively; p ϭ 0.045). The ability of the W174A mutation to suppress invasion implies that HP1 Hs␣ interactions with Current models for heterochromatin formation and spreading include binding of HP1 to meK9 H3, recruitment of SUV39h histone methyltransferases through direct interactions with HP1, and subsequent methylation of adjacent histones (17,20). Thus, there is a potential that the decrease in invasion observed after expression of EGFP-HP1 Hs␣ could result in changes in meK9 H3. Western analysis and immunolocalization were performed to determine whether the localization and abundance of di-or tri-meK9 H3 changed after infection of the adenoviral constructs in MDA-MB-231 cells. Neither gross localization (Fig. 4A) nor expression levels (Fig. 4B) of di-or tri-meK9 H3 were observed to change after expression of EGFP-NLS, EGFP-HP1 Hs␣ , EGFP-I165E, or EGFP-W174A. These observations lead to the conclusion that global changes in di-and tri-meK9 H3 are unlikely to be involved in the suppression of the invasive potential of the metastatic breast cancer cell line, MDA-MB-231, by EGFP-HP1 Hs␣ .
HP1 Hs␣ Does Not Affect Growth of MDA-MB-231 Cells-Because HP1 Hs␣ is able to suppress invasion, a key step in metastasis, and is down-regulated in metastatic tissue from breast cancer patients, HP1 Hs␣ is hypothesized to be a metastasis suppressor protein (9). Metastasis suppressor proteins are defined as being able to suppress metastasis without affecting the growth of tumor cells (3). To test for the effect of HP1 Hs␣ expression on the growth rate of MDA-MB-231 cells, growth curves of uninfected cells and of cells infected with one of the adenoviral constructs (EGFP-NLS, EGFP-HP1 Hs␣ , EGFP-W174A, and EGFP-I165E) were compared (Fig. 5A). During the time period when the in vitro invasion assays were performed (48 -72 h post-infection), there was no difference in growth between the samples. Therefore, growth rate differences do not explain the decrease in invasion of MDA-MB-231 cells infected with EGFP-HP1 Hs␣ . After 72 h, all samples infected with adenovirus showed a slower growth rate than uninfected cells. No statistical difference was observed among the EGFP-NLS-, EGFP-HP1 Hs␣ -, EGFP-I165E-, and EGFP-W174A-infected cells through 144 h post-infection (Fig. 5A). The slower growth rate is most likely because of adenoviral effects.
To further determine the effects of HP1 Hs␣ expression in MDA-MB-231 cells, cell cycle progression was investigated using propidium iodide fluorescence-activated cell sorting. No differences were observed between uninfected cells and those infected with EGFP-HP1 Hs␣ during G 0 -G 1 , S, or G 2 -M phases of the cell cycle 48 h post-infection (Fig. 5B). Although there were no changes during G 0 -G 1 , S, or G 2 -M phases, there might be differences in the ability of the cells to undergo DNA replication. BrdUrd incorporation studies were performed using uninfected MDA-MB-231 cells, cells expressing EGFP-NLS, and cells expressing EGFP-HP1 Hs␣ . No differences in BrdUrd staining were observed between the samples after 12, 6, 3, or 1 h of BrdUrd treatment (84 h post-infection) ( Fig. 5C and data not shown). Taken together, expression of EGFP-HP1 Hs␣ in MDA-MB-231 cells does not result in altered DNA replication, cell cycle progression, or cellular growth rate. These data are consistent with the hypothesis that HP1 Hs␣ is a metastasis suppressor in breast cancer cells.
Knock-down of HP1 Hs␣ in MCF-7 Cells-If expression of HP1 Hs␣ is sufficient to suppress invasive potential in breast cancer cells, a reduction of HP1 Hs␣ is predicted to increase the invasive ability of poorly invasive breast cancer cells. To test this prediction, the poorly invasive nonmetastatic breast cancer cell line, MCF-7, previously shown to have high levels of HP1 Hs␣ (9), was investigated for phenotypic differences after knock-down of HP1 Hs␣ by RNAi. MCF-7 cells were infected with adenovirus expressing short hairpin RNAi molecules targeted to either HP1 Hs␣ (shHP1 Hs␣ ) or GFP (shGFP) as a negative control, and expression levels of HP1 Hs␣ were subsequently compared. Levels of HP1 Hs␣ decreased to less than 5% of that present in uninfected cells 48 h post-infection and remained at that level for ϳ7 days (Fig. 6A and data not shown). As there are three mammalian HP1 family members, the effects of HP1 Hs␣ knock-down on the expression of HP1 Hs␤ and HP1 Hs␥ were investigated. No dramatic changes in protein levels of HP1 Hs␤ or HP1 Hs␥ were detected upon knock-down of HP1 Hs␣ (Fig. 6A). These data show that HP1 Hs␣ knock-down can be achieved through adenoviral shRNA infection of MCF-7 cells and that the loss of HP1 Hs␣ does not dramatically alter the levels of HP1 Hs␤ or HP1 Hs␥ .
Previous immunofluorescent data have shown that HP1 Hs␣ is organized in multiple discrete foci in MCF-7 cell nuclei (9). Uninfected cells showed similar immunofluorescent signal intensity and punctate distribution of HP1 Hs␣ in the nucleus, as reported previously (9). Cells infected with shGFP as a negative control showed an intensity of signal and localization of HP1 Hs␣ similar to uninfected cells. The localization of HP1 Hs␣ was investigated to determine the effects of knock-down in MCF-7 cells using an EGFP marker in the shHP1 Hs␣ viral backbone to discriminate between infected and uninfected cells (Fig. 6B). Infection efficiency based on visualization of EGFP fluorescence was ϳ90%. Cells infected with shHP1 Hs␣ were evaluated based on expression of GFP. Those cells not expressing EGFP showed a pattern of HP1 Hs␣ expression similar to uninfected cells (Fig. 6B). In contrast, cells with high EGFP levels showed an almost complete loss of HP1 Hs␣ immunofluorescent signal, although the signal that remained was present in a punctate pattern (Fig. 6B). Therefore, the levels of HP1 Hs␣ were reduced, but the localization pattern of remaining HP1 Hs␣ continued to be in discrete foci.

Knock-down of HP1 Hs␣ in MCF-7 Cells Results in Increased Invasive
Potential-To test for whether reduction of HP1 Hs␣ in a poorly invasive nonmetastatic cell line increases invasion, an in vitro assay was used to evaluate the invasive ability of MCF-7 cells with or without knockdown. Cells infected with a virus expressing shGFP as a negative control did not have a significantly altered invasive potential compared with uninfected cells (87 versus 100%, respectively; p ϭ 0.338) (Fig. 7A). In  . Twelve independent cultures were tested for the uninfected cells, the cells infected with EGFP-NLS, and the cells infected with EGFP-HP1 Hs␣ ; 10 independent cultures were tested for the cells infected with I165E; and 8 independent cultures were tested for the cells infected with W174A. B, graph representing MDA-MB-231 cells either uninfected or infected with a virus expressing EGFP-HP1 Hs␣ analyzed for cell cycle stage using propidium iodide fluorescence-activated cell sorting analysis. Two independent cultures were tested. C, graph representing BrdUrd incorporation versus DAPI positive cells in MDA-MB-231 cells either uninfected or infected with an adenovirus expressing EGFP-NLS or EGFP-HP1 Hs␣ . Three independent cultures were tested. The p values were calculated using Excel t test.
contrast, cells expressing shHP1 Hs␣ increased the invasiveness of MCF-7 cells compared with shGFP (136 versus 87%, respectively; p ϭ 0.028) or compared with uninfected cells (136 versus 100%, respectively; p ϭ 0.017) (Fig. 7A). These results provide evidence that a reduction of HP1 Hs␣ is sufficient to increase the invasive potential of MCF-7 breast cancer cells.
Knock-down of HP1 Hs␣ Does Not Alter Growth of MCF-7 Cells-To further explore the possibility that HP1 Hs␣ is a metastasis suppressor, the effects of HP1 Hs␣ knock-down on growth in MCF-7 cells was investigated. The growth rates of MCF-7 cells uninfected and infected with either shHP1 Hs␣ or shGFP were compared. No changes in growth between the different samples were noted during the time points when invasion was assayed (Fig. 7B). Therefore, growth rate differences do not explain the increase in invasion of MCF-7 cells. However, the adenovirus-infected cells showed a reduced growth rate compared with uninfected cells after 72 h post-infection, most likely because of viral effects (Fig. 7B). Thus, studies in both highly invasive metastatic and poorly invasive nonmetastatic breast cancer cells show that modulation of HP1 Hs␣ levels changes the invasive properties of the cells without affecting cellular growth rates.

DISCUSSION
Cancer progression is accompanied by cumulative genetic alterations, with transformation and immortalization of cells being well studied (35). However, the molecular mechanisms that result in dissemination of cells from the primary tumor to distant sites in the body are poorly understood (3, 36 -40). Although progress is being made in understanding stromal interactions, motility/chemoattractant migration, and the epithelial to mesenchymal transition that accompanies metastasis, there is still a deficiency in mechanistic knowledge and predictive power in treating breast cancer metastasis (41)(42)(43). Because of this lack of understanding and an inability to accurately anticipate which patients will develop metastatic lesions, many patients who will never develop metastatic disease are treated unnecessarily with cytotoxic therapeutic agents (5,44,45). In other cases, those who will acquire metastases are often either undertreated or given therapies to which the metastases are resistant (5,44,45). This lack of diagnostic power may account for the large difference in 5-year relative survival rates between localized and metastatic breast cancer (98 versus 26%, respectively) (46). Better prognostic and predictive markers are needed to correctly assess and treat patients.
There is precedent for regulating cancer progression through chromatin organization (47). Alterations in global chromatin structure could simultaneously affect disparate pathways involved in cancer progression. The HP1 family of proteins, which regulates chromatin organization, has been shown to control chromosomal stability and transcriptional regulation (24, 48 -51). Improper chromatin organization, including abnormal expression and localization of HP1 family members, is linked to leukemia cell proliferation (52). Additionally, HP1 Hs␣ has been associated with tumor progression in multiple human cancers; HP1 Hs␣ mRNA levels are reduced in advanced forms of papillary thyroid carcinoma, medulloblastoma, and breast cancer (9,53,54).
The cause of HP1 Hs␣ mRNA down-regulation in MDA-MB-231 cells relative to MCF-7 cells has been investigated, with evidence showing differential regulation by an E-box element in the 5Ј-promoter region (30). Yet, how HP1 Hs␣ mediates a change in invasive potential has not  been elucidated. We hypothesize that HP1 Hs␣ regulates genes involved in invasion/metastasis in breast cancer cells. This could be dependent upon the ability of HP1 Hs␣ to dimerize and/or interact with partner proteins to regulate gene expression.
In breast cancer, decreased levels of HP1 Hs␣ have previously been shown to correlate with metastasis to distant sites (9). We have demonstrated that the level of HP1 Hs␣ regulates in vitro invasion; increased expression of HP1 Hs␣ reduces invasive potential, whereas decreased expression enhances invasive potential. Altering levels of HP1 Hs␣ does not completely reverse the invasive phenotype. This incomplete effect is likely due to cell-by-cell variation in expression of HP1 Hs␣ following viral infection. A prior post-invasion analysis of the cells determined that the majority of the MDA-MB-231 cells (95%) that transverse the membrane in the invasion chamber have reduced GFP fluorescence compared with cells that remain in the upper portion of the chamber (9). Similarly, only a 50% increase in invasion relative to controls was observed upon knock-down of HP1 Hs␣ in MCF-7 cells. This limited increase may be due to residual levels of HP1 Hs␣ following knock-down. Alternatively, redundant functions of HP1 Hs␤ and/or HP1 Hs␥ with HP1 Hs␣ could also account for partial effects.
There are several potential explanations that could account for the alterations in invasion observed upon modulation of HP1 Hs␣ levels. First, cellular adhesion could change the ability of cells to associate with the membrane for initiation of invasion. However, adhesion assays show no difference between experimental and control cells. 4 Second, changes in H3 K9 methylation status might occur upon variation in HP1 Hs␣ function leading to alterations in gene expression related to invasion. We did not observe global changes in localization or expression levels of di-and tri-meK9 H3 in metastatic MDA-MB-231 breast cancer cells after introduction of wild type and mutant forms of EGFP-HP1 Hs␣ . Further analysis is required to determine whether changes in meK9 H3 occur at specific loci regulated by HP1 Hs␣ . Third, changes in growth rate could account for differential invasive potential. However, HP1 Hs␣ regulation of invasion is not accompanied by changes in cellular growth, because growth rate, cell cycle stage, and DNA replication were unchanged among uninfected, control, and experimental cells (Fig. 5). This is in contrast to reports in other cellular systems where modulation of HP1 levels alters cell cycle progression (55,56). Differences could reflect tissue, developmental, and/or species-specific roles of HP1 (57).
In breast cancer cells, HP1 Hs␣ possesses characteristics of a metastasis suppressor: limiting invasion while not altering growth rate. Clinical studies are consistent with these in vitro findings, where down-regulation of HP1 Hs␣ is found in metastatic breast cancer tissue but not in primary breast cancer tissue (9). To provide further evidence that HP1 Hs␣ is a metastasis suppressor, in vivo metastasis assays using a mouse model are necessary (58).
To dissect the mechanism of how HP1 Hs␣ regulates invasion of breast cancer cells, two functional properties of the HP1 Hs␣ CSD were analyzed: dimerization of HP1 and interactions with PXVXL motif proteins (21,22). Here, two CSD mutations were investigated for effects on breast cancer cell invasion. The HP1 Hs␣ I165E mutation disrupts CSD dimerization and, consequently, interactions with protein partners that require dimerization (20,33,34). The analogous mutation in mouse HP1␤ (I161E) causes diffuse HP1␤ staining throughout the nucleus (22). In addition, the mutation in Drosophila HP1 (I191E) causes a reduction of HP1 at euchromatic sites on polytene chromosomes. 5 The HP1 Hs␣ W174A mutation disrupts PXVXL-containing partner interac-tions, without disrupting dimerization (20,22,33). The analogous mutation in mouse HP1␤ (W170A) does not result in mislocalization, showing anticipated nuclear foci (22). The analogous mutation in Drosophila HP1 (W200A) retains binding at euchromatic sites. 5 Consistent with these findings, the I165E mutation shows diffuse nuclear staining, whereas the W174A mutation shows punctate nuclear foci in human breast cancer cells (Fig. 3). Collectively, these studies demonstrate that the majority of HP1 chromosomal associations require dimerization rather than interactions with PXVXL-containing partner proteins.
The I165E and W174A point mutations also differentially affect in vitro invasion in MDA-MB-231 cells. The I165E mutant protein is unable to suppress invasion, whereas the W174A can suppress in vitro invasion. This suggests that dimerization is also required for suppression of breast cancer in vitro invasion, whereas interactions with PXVXL partner proteins are not required.
The necessity of having dimerization, but not the PXVXL binding platform, for suppressing breast cancer cell invasion does not eliminate the possibility of partner protein involvement. Although many proteins interact with the HP1 family members through the PXVXL motif, others do not use this mechanism. For example, interaction between BRG1 and HP1 Hs␣ does not require the PXVXL platform (25). In addition, the interaction of HP1 family members with SUV39h1 does not involve a PXVXL motif but does require HP1 dimerization (26). Thus, HP1 partner proteins that do not interact through the PXVXL platform might also play a role in HP1 Hs␣ -mediated regulation of breast cancer invasion.
Another possible mechanism for regulating breast cancer invasion is HP1 Hs␣ dimerization itself. Chromosomal association of HP1 generates a compact chromatin structure, possibly because of interactions between HP1 proteins. The loss of this compact structure could lead to inappropriate activation of genes that are responsible for the invasive phenotype of breast cancer cells (24,48,59). In addition, dimerization of HP1 molecules at distant chromosome sites is proposed to form loop structures that bring enhancers into proximity with promoters (24,60). In such a case, reduction in HP1 levels would lead to repression of genes that prevent invasion. In mammalian cells, HP1 dimerization is complex and might involve both homo-and heterodimerization (61). The composition of a heterodimer could specify chromatin localization and/or partner interactions. This idea is consistent with data showing partial overlapping localization of HP1 family members (62)(63)(64). Further studies that would distinguish the functions of HP1 homo-and heterodimers would help to clarify this issue. We propose a model in which HP1 Hs␣ regulates genes involved in breast cancer cell invasion. Evidence that HP1 Hs␣ regulates genes comes from microarray studies in Drosophila and mammalian cells, where hundreds of genes both increase and decrease in expression upon a reduction in HP1 levels (56,65). Therefore, it would be beneficial to identify genes in which expression levels depend on HP1 Hs␣ dimerization and to evaluate the role they play in invasion and metastasis. These genes, and the pathways in which they are involved, would offer new insight into the process of metastasis and possibly provide new markers for diagnosis and prognosis.