Maintenance of integrated proviral gene expression requires Brm, a catalytic subunit of SWI/SNF complex

We show here that MuLV-based retrovirus vector transgene expression is rapidly silenced in human tumor cell lines lacking expression of Brm, a catalytic subunit of the SWI/SNF chromatin remodeling complex, even though these vectors can successfully enter, integrate, and initiate transcription. We detected this gene silencing as a reduction in the ratio of cells expressing the exogenous gene rather than a reduction in the average expression levels, indicating that down-regulation occurs in an all-or-none manner. was protected from silencing and maintained in Brm-deficient host cells by exogenous expression of Brm but not BRG1, an alternative ATPase subunit in the SWI/SNF complex.


INTRODUCTION
Retroviruses are known to integrate into host cell chromosomes as proviruses and to express viral genes even after host cell proliferation. One problem limiting development of retroviral vectors with long-term expression, however, is gene silencing (1). DNA methylation has been reported as the major epigenetic DNA modification that occurs in conjugation with provirus down-regulation (2)(3)(4) and serves as a signal for association with methyl-CpG-binding protein 2 (MeCP-2), which together form large protein complexes containing histone deacetylases (HDACs) (5). Therefore, silenced retroviral genes have been sometimes reactivated by treating cells with either the DNA methyltransferase inhibitors, 5-azacytidine (5-aza-C) and 5-azadeoxycytidine (5-azadC) or HDAC inhibitors, such as trichostatin A (TSA). In many cell types that exhibit gene silencing, however, it is not clear whether provirus methylation is the primary cause or simply reflects transcriptional repression. Gene silencing that is not mediated by DNA methylation has been suggested (6-9).
The proteins required to maintain a repressed state are Pc-G complex, whereas those required for persistence of expression are trx-G complex. Pc-G and trx-G protein complexes do not contribute to the initiation of specific target genes but instead counteract each other to repair previously established chromosomal domains of specific genes throughout development (10).
Insight into the role of the trx-G protein complex in transcriptional regulation has came from studies of the Drosophila Brahma gene and its mammalian homologues Brm and BRG1. The products of these genes have DNA-dependent ATPase activity and are classified as SWI2/SNF2 family proteins. The mammalian SWI/SNF chromatin remodeling complex contains either Brm or BRG1 but not both (11). The differences in biochemical function between Brm and BRG1 are largely unknown, however we observed the Brm or BRG1 subunit has distinct target specificity to facilitate gene activation through AP-1 (12). These findings provide mechanistic links between epigenetic transcriptional regulation and chromatin remodeling.
We herein describe retroviral gene silencing, which occurs very rapidly in a discontinuous and stochastic manner, in certain human tumor cell lines, and we show that

Rapid Retroviral Gene Silencing in Human Brm-Deficient Tumor Cell Lines--We
previously reported that transduction efficiency by VSV-G pseudotyped vector is quite high in cell lines originated from human solid tumors as well as murine fibroblasts (17,18). To closely examine why a significant cellular population in these C33A and SW13 cultures failed to express LacZ, these cells as well as MDA-MB435 (as positive control) were transduced with LacZ virus at a low MOI (about 0.4) to minimize the introduction of multiple proviral copies into a single cell. The transduced cell cultures were grown for 1 day to complete the retroviral integration and then, seeded at a low cellular concentration for single colony formation. The colonies formed 3 days after seeding were stained for LacZ to assess LacZ expression. An MDA-MB435 cell that was transduced with LacZ virus formed a colony in which all the progeny cells express lacZ (Fig. 1A). We define such colony as a "positive colony" (Fig. 1B). The cell that escaped from the transduction formed a colony that is exclusively composed of expression-negative cells. We define such colony as a "negative colony" (Fig. 1B). Surprisingly, most of the colonies formed by LacZ-virus transduced C33A and SW13 cells were composed of mixed populations of positive and negative cells (Fig. 1A). Such a colony was defined as a "mosaic colony" To semi-quantify the extent of gene silencing, we tentatively defined "mosaic colony ratio" of a seeded culture after the transduction by dividing the number of mosaic colony by the sum of the number of positive colony and the number of mosaic colony.
We determined the mosaic colony ratios for C33A, SW13, and MDA-MB435 and several other human tumor cell lines as well as murine fibroblast cell lines using the same LacZ virus. In most cell lines examined, including MDA-MB435, H1299, and SW620, HeLa S3 and 3Y1, the mosaic colony ratios were low and ranged from 0.02 to 0.15. C33A and SW13 cells, however, exhibited a high mosaic colony ratio ( Fig. 2A). Two other cell lines, Saos2 and G401, showed similar properties to C33A and SW13 cells, although the mosaic colony ratio was lower.
Comparison of gene expression patterns among these cell lines might reveal host factors that inhibit or accelerate retroviral gene silencing. We were interested in Brm and BRG1, which are the essential subunits of SWI/SNF chromatin remodeling complex and have DNA-dependent ATPase activity, as candidates for inhibitory factors. Expression of both proteins is reported to be absent or very low in both C33A and SW13 cells (19)(20)(21)(22).
Therefore, we screened the cell lines listed in Fig. 2A for Brm and BRG1 expression. Saos2 cells were previously reported to produce Brm by Western blotting analysis with a monoclonal antibody raised against Brm (21), and we were able to reproduce this finding.
Since we found this monoclonal antibody to be weakly cross-reactive with BRG1, we used a non-cross-reactive goat polyclonal antibody and confirmed that these cell lines expressed no detectable Brm. Therefore the rapid retroviral gene silencing appears to be correlated with the absence of endogenous Brm expression ( Fig. 2A), whereas loss of BRG1 does not appear to be related directly to retroviral gene silencing. caused no changes in the percentage of either "mosaic colony" or "positive colony" as was shown in SW13 cells. These findings suggest that histone acetylation is involved in retroviral gene silencing in Brm-deficient cell lines and that the process does not involve a mechanism mediated through CpG methylation. Considering the possibility that SWI/SNF complex counteracts the Pc-G protein complex to suppress gene silencing, we next examined whether recruitment of Pc-G protein complexes to the LTR is modulated by Brm-containing SWI/SNF complex. Among mammalian Pc-G, only YY1 is currently known to have specific DNA-binding activity, and an YY1 binding site is present in the 5'-region of the MuLV-LTR (28) (Fig. 5). By

Brm-Containing SWI/SNF Complex Inhibits Recruitment of YY1 Complex and Enhances Acetylation of Specific Lysine Residues in Histone H4 in the 5'LTR Region--Since
ChIP-PCR assay using the same cellular extracts described above, we examined whether the amount of YY1 in the 5'LTR is reduced by the expression of Brm. In the immunoprecipitates with anti-YY1 antibody, there was a lower amount of 5'LTR sequence in C33A cells transduced with Brm virus compared with C33A cells transduced with control virus.
Since HDAC1 and HDAC2 have been reported to form large Pc-G protein complexes with YY1 (29,30), we next analyzed these two proteins in the 5'LTR. In Brm virus-transduced C33A cells, the amounts of HDAC1 and HDAC2 were also reduced.
Since the PCR product of the primer pair 1 + 2 does not include the YY1 recognition site, another primer pair 3 + 4 was used to cover the YY1 recognition site, although it can not distinguish the 5'-LTR and 3'-LTR. The results were similar to those obtained with the primer pair 1 + 2 (Fig.5) inactive chromatine, also reduced by the introduction of Brm (Fig.5). These data are consistent with the notion that when the provirus MuLV-LTR integrates into C33A cells, Pc-G protein complexes including YY1, HDAC1, and HDAC2, are efficiently recruited to silence retroviral gene expression by deacethylating specific lysine residues in histone H4 and this suppressive effect can be efficiently counteracted by functional SWI/SNF complex.
In the U3 region of the   Figure 1B. In C33A and SW13 cells, the mosaic colony ratios were reduced when cells were to transduce with the LacZ virus at higher MOIs (unpublished observation), suggesting that proviral copy number rather than cellular physiology determines the degree of transcriptional down-regulation. Therefore, we believe the variegated gene silencing observed here reflects drastic structural changes that occurred around the proviral LTR and that these structural changes are the all-or-none result of the opposing actions of trx-G and Pc-G protein complexes.
Considering the fact that the retroviral gene silencing observed in Brm deficient cell lines was efficiently suppressed by Brm introduction (Fig. 3), it is worth pointing out that the silencing was not fully recovered by the treatment with the HDAC inhibitor, CHAP31 alone (Fig.4). This might suggest the possibility that Brm have some additional biochemical roles other than counteracting HDAC activity as a component of SWI/SNF complex in the chromosome (Fig.5).
The search for such biochemical functions of Brm, an ATPase with a putative helicase motif, would be important.
Retrovirus vectors have been designed recently to overcome chromosomal position effects (34,35). For example, the chicken hypersensitive site 4 (cHS4) of the chicken globin LCR, which acts as an insulator, was cloned into MuLV-LTR. In some cell lines, the probability of integrated proviral gene expression increased, and the level of However, the transduced SW13 (or C33A) formed colonies that are composed of both transgene-expressing cells and transgene-non-expressing cells. We defined this colony as a "mosaic colony". We hypothesized that in SW13 or C33A, the transgene expression was silenced in a discontinuous and stochastic manner.