K562 cells implicate increased chromatin accessibility in Alu transcriptional activation

Abstract

Alu repeats in K562 cells are unusually hypomethylated and far more actively transcribed than those in other human cell lines and somatic tissues. Also, the level of Alu RNA in K562 cells is relatively insensitive to cell stresses, namely heat shock, adenovirus infection and treatment with cycloheximide, which increase the abundance of Alu RNA in HeLa and 293 cells. Recent advances in understanding the interactions between DNA methylation, transcriptional activation and chromatin conformation reveal reasons for the constitutively high level of Alu expression in K562 cells. Methylation represses transcription of transiently transfected Alu templates in all cell lines tested but cell stresses do not relieve this repression suggesting that they activate Alu transcription through another pathway. A relatively large fraction of the Alus within K562 chromatin is accessible to restriction enzyme cleavage and cell stresses increase the chromatin accessibility of Alus in HeLa and 293 cells. Cell stress evidently activates Alu transcription by rapidly remodeling chromatin to recruit additional templates.

INTRODUCTION

Alu repeats contain internal promoter elements for RNA polymerase III (Pol III) (1). The presence of one million Alus distributed throughout the human genome therefore creates a tremendous transcriptional potential. Despite this potential, the steady state abundance of Pol III directed primary Alu transcripts [named full length (fl) Alu RNA] is usually very low in tissues and cultured cells (2–4).

Various cellular ‘stresses’, including heat shock, viral infection and cycloheximide exposure, increase the abundance of flAlu RNA (5–8). For example, the level of flAlu RNA increases within 20 min of exposing cells to cycloheximide (7,8). The rapidity of this response to a translational inhibitor suggests that it results from a modification of existing factors. These same cell insults dramatically increase the abundance of SINE (short interspersed elements) RNAs in other mammals and in silkworm showing that this response is evolutionarily conserved (8–14). Also, the SINE stress response occurs in living animals and is subject to tissue specific regulation (15). The stress-induced increases appear to be specific for SINEs, as the same stresses do not activate other Pol III directed genes (reviewed in 16); also, cycloheximide and long-term viral infection actually inhibit Pol III transcription (17,18). After heat shock, the maximum levels of SINE and heat shock protein RNAs occur at about the same time (8,10,13,15). SINE RNA may have a defined role in regulating protein synthesis, especially under cell stress conditions (19). While the exact function of these transcripts remains to be proven, SINEs are expressed in a tissue-specific manner and behave like classic heat shock genes in response to stress.

Since Alus are conceptually the dispersed members of a vast multigene family, an intriguing question is how so many ‘genes’ are regulated and stress induced. Accumulating evidence indicates that regulated Alu expression is determined on many levels some of which act locally upon individual members of the family (1,20–23). In contrast, DNA methylation and chromatin might each globally repress transcription of this repetitive sequence family and therefore potentially direct the regulation of the Alu stress response.

Methylated cytosine (5-meC) in mammalian DNAs occurs exclusively within CpG dinucleotides (24). Approximately one-third of the CpGs within human DNA reside within Alu elements which are almost completely methylated in adult somatic tissues (25–28). As an exception to the general methylation of Alus, a majority of an evolutionarily young (Y), i.e. CpG rich, Alu subfamily is unmethylated in male germ line tissues (26–28). Older Alu subfamilies are also hypomethylated in the male germ line; however, their relative level of methylation has been technically more difficult to assess (26–28). The divergence of the older Alus decreases the fidelity of their consensus restriction sites and hybridization to subfamily specific probes.

DNA methylation represses Pol III transcription in vivo of tRNA and adenovirus VAI RNA genes (29,30). 5-meC binding proteins repress Pol II directed transcription by recruiting histone deacetylases to condense chromatin (31–33). This same mechanism probably applies to Pol III transcription, as a repressor which is specific for 5-meC inhibits Alu transcription in vitro (34). Since histones are normally deacetylated (35), Alu hypermethylation might merely reinforce this default chromatin structure and transcriptional repression (36).

Chromatin represses SINE transcription (12,37–39). Furthermore, viral infection increases the template activity of SINE chromatin in vitro (12,39). These functional assays suggest that an opening of chromatin structure accompanies SINE transcriptional activation. As one possibility, specific binding factors might regulate Alu chromatin structure and transcription (40). In this case, heterologous SINEs in other species would presumably also be similarly regulated by factors which are specific for each of these distinct sequences. As a more general alternative, viral infection increases the activity of a basic transcription factor, TFIIIC2, and this increase has also been proposed to activate SINE transcription (5,6,12,39,41,42; reviewed in 7,16).

In summary, Alu transcription might be globally increased by activating TFIIIC or by derepressing the inhibition which either DNA methylation or chromatin structure imposes. However, the effects of TFIIIC activation, methylation and chromatin structure are interrelated since activated TFIIIC2 has histone acetyltransferase activity (43,44).

Alus are found to be unusually hypomethylated in human K562 cells, an erythroleukemic line. This and other unusual properties of Alus in K562 cells provide an opportunity to investigate the effects that methylation and other potential regulators have on Alu transcription.

MATERIALS AND METHODS

Cell culture

HeLa, 293 and K562 cells were maintained in DMEM, αMEM, or RPMI1640 media with 10% new-born calf serum. For heat shock, cells were incubated at 45°C for 30 min and recovered at 37°C for the times indicated (8). For cycloheximide, cells were exposed to 100 µg/ml. Viral infection was performed using adenovirus type 2 (19), and VAI RNA was monitored to verify viral infection (data not shown). To determine RNA lifetimes, cells were treated with actinomycin D (5 µg/ml) (19).

The flAlu RNA expression plasmid XAT was transfected using calcium phosphate (19) or DOTAP liposomal transfection reagent (Roche, Indianapolis, IN). RNA was isolated 48 h after transfection. Plasmids were methylated with SssI CpG specific methylase (New England Biolabs, Beverly, MA) and complete methylation was confirmed by HpaII digestion.

RNA analysis

Cytoplasmic RNA was isolated as described by Liu et al. (2). Primer extension with reverse transcriptase and the indicated primers was used to assay flAlu RNA (GCGATCTCGGCTCACTGCAAG), VA1 RNA (AAAAGGAGCGCTCCCCCGTT), 7SL RNA (ATGCCGAACTTAGTGCGG) and β-actin mRNA (GCCTGGGGCGCCCCACGAT) (2,19). All products had the expected lengths as assayed by PAGE (data not always shown).

For cDNA cloning, primer extension products of flAlu RNA from K562 cells were size selected by PAGE. This first strand cDNA was PCR amplified using either AAAGAATTCGGCCGGGCGCGGTGGCT or GCCTGTAATCCCAGCACTTTGGA as second strand primers in conjunction with an opposing primer, ACGCCATTCTCCTGCCTCAG. PCR products were cloned using the TOPO TA (Invitrogen, Carlsbad, CA) method.

For northern analysis, 60 µg of cytoplasmic RNA was separated on a 5% denaturing polyacrylamide gel and transferred to a nylon membrane (Hybond-N+, Amersham, Piscataway, NJ). Radiolabeled oligonucleotide TCACCGTTTTAGCCGGGATGGT was used to detect Alu transcripts (19).

Nuclear run-on analysis was performed with equal numbers (2 × 10⁷) of isolated nuclei from HeLa and K562 cells in the presence of 250 µCi of [α-³²P]UTP and 10 µg/ml of α-amanitin (45). Radiolabeled RNA was isolated and hybridized to nitrocellulose immobilized plasmids specific for Alu, 5S rRNA, 18S rRNA and pUC19, a negative control.

Chromatin accessibility

Cells were rinsed in PBS and pelleted by centrifugation. The pellet was resuspended in buffer D (5 mM PIPES, pH 8, 85 mM KCl, 1 mM CaCl₂, 5% sucrose) with 0.5% Nonidet P-40 (46). Cells were left on ice for 10 min during which lysis occurred. The nuclei were rinsed three times in buffer D (46).

For restriction digestion, nuclei (4 × 10⁶ cells) were suspended in buffer #2 (New England Biolabs). After addition of 20 U of either HaeIII, HinfI or both, nuclei were incubated at 37°C for 2 h. Digestion was stopped by adding an equal volume of lysis buffer [50 mM Tris–HCl, pH 7.5, 15 mM NaCl, 15 mM EDTA, 0.3% sodium dodecyl sulfate (SDS)]. The sample was treated for 1 h at 37°C with 50 µg/ml RNase A, overnight at 56°C with 50 µg/ml proteinase K, and DNA was isolated.

For Southern blots, gels (1.2% agarose) were transferred in 0.4 M NaOH to Hybond-N+ membrane (Amersham). After drying, the membrane was incubated with prehybridization buffer (5× SSPE, 5× Denhardt’s solution, 0.5% SDS and 100 µg/ml sheared salmon sperm DNA) at 52°C and then hybridized with a Ya5 Alu subfamily-specific oligo probe (PV 51) (28,47) or an α-satellite probe, TGCATTCAACTCACAGAGTTGAAC (48). Membranes were washed twice with 5× SSPE, 0.1% SDS at room temperature and for 5 min at 52°C.

Restriction assays of Alu methylation

Using Southern analysis, several tests of Alu methylation have been performed (Fig. 1 and not shown). To test the methylation of a BstUI site, DNA (∼20 µg) was predigested by Tth111 and reisolated by phenol extraction and ethanol precipitation (28). These samples were divided into two equal volume aliquots which were digested for 4 h with either BstUI or TaqI using twice the units required for 1 h (28). Alternatively, consensus HaeIII–HinfI Alu fragments were size-selected by gel electrophoresis and tested for methylation using HpaII and MspI and SmaI and XmaI (25). Southern analysis of the digests was performed using oligonucleotide PV 51 and hybridization conditions that select for the Ya and Ya5 subfamilies as well as a cloned probe and conditions that are not subfamily specific (25,28).

Figure 1.

Southern analysis of Alu methylation. Samples of HeLa, K562 and 293 cell DNAs (lanes indicated) were predigested with Tth111 and each divided into two equal aliquots which were next digested with either BstUI (lanes B) or TaqI (lanes T). Southern blot analysis of these samples was performed under conditions that select for the Ya5 Alu subfamily (28).

Data analysis

Gels were analyzed with a Fuji BAS 1000 phosphorimager or Fuji blue X-ray film with an intensifier screen. BAS 1000 images were quantified with Molecular Dynamics Storm 860 hardware and Molecular Dynamics Image Quant software.

RESULTS

Hypomethylation of K562 Alus

Preliminary studies (not shown) indicated that K562 cells have an unusually low level of Alu methylation approximating that observed in sperm DNA (28). This raised the possibility of comparing Alu transcription in cell lines having different degrees of methylation. As one relative measure of Alu methylation, we tested the cleavage of a consensus BstUI site by blot hybridization (Fig. 1). The BstU site (CGCG) compounds four methylation sites making it especially sensitive to Alu methylation (27,28). To compare the amount of DNA in these samples, an equivalent aliquot was digested with TaqI (Fig. 1). To target conserved restriction sites (Introduction), the hybridization conditions employed in this experiment select for the young Ya5 Alu subfamily (28). Cleavage of 293 cell Alus by BstUI is undetectable, indicating the nearly complete methylation of this site (Fig. 1). In contrast, a small fraction of the HeLa and a larger fraction (∼5-fold) of K562 cell Ya5 subfamily Alus are cleaved by BstUI (Fig. 1).

This experiment interrogates the methylation of a single restriction site in a single Alu subfamily. We further observe that other sites in K562 cell Alus are also hypomethylated and that hypomethylation is not restricted to the Ya5 subfamily but extends to the larger Y subfamily (data not shown). The Ya5 subfamily accounts for only 0.1% of all Alus whereas the larger Y subfamily accounts for 10% (49). As discussed below, members of the young Alu subfamilies appear to be more abundantly expressed than older Alus.

Unusually high expression of flAlu RNA

Primer extension analysis was used to assay for Pol III-directed Alu transcripts in K562 cells (Fig. 2). The level of flAlu RNA in K562 cells is extremely high as compared to HeLa or 293 cells (Fig. 2, compare lanes C). One Alu locus is known to be expressed in K562 cells (50). Four cDNA clones of K562 Alus were identical but 23 contained distinct sequences (GenBank AF205194–AF205217) indicating that many Alu loci are transcribed (2–4). Also, 14 of the 24 distinct cDNA sequences belong to young subfamilies (namely the Ya, Ya5 and Ya8 subfamilies) which together constitute only about 10% of all Alus (47,49). As reported above, young subfamiles are hypomethylated in K562 cells but younger Alus also tend to be more actively expressed (2–4).

Figure 2.

Primer extension analysis of flAlu RNA. RNA from HeLa, 293 or K562 cells that had been treated as described below was analyzed by primer extension with reverse transcriptase. ‘C’ refers to untreated controls in all experiments. Primers for either β-actin mRNA or 7SL RNA (not shown) were included in all primer extension reactions to control for loading. (A) Cells were infected with adenovirus ‘Adv’; primer extension of VAI RNA verifies viral infection in all cases (data not shown). The decrease in β-actin mRNA in infected 293 cells is probably a consequence of infection as equal amounts of RNA are used in all lanes. (B) Cells were heat shocked and allowed to recover for the number of hours indicated. (C) Cells were exposed to cycloheximide for the number of hours indicated.

We examined the effects of cellular insults upon flAlu RNA in K562 cells. Viral infection has no effect upon flAlu RNA in K562 cells but increases flAlu RNA in HeLa and 293 cells to levels approximating that in K562 cells (Fig. 2A). Heat shock and cycloheximide cause slight increases in flAlu RNA in K562 cells (Fig. 2B and C). After 4 h of heat shock recovery, flAlu increases 8-fold in HeLa cells; the maximum increase in K562 is ∼2-fold (Fig. 2B). The level of flAlu RNA in HeLa cells treated with cycloheximide exceeds that in control K562 cells but cycloheximide causes only a modest (∼2-fold) increase in flAlu RNA in K562 cells (Fig. 2C). In essence, flAlu RNA is constitutively expressed at high levels in K562 cells.

K562 Alus are more actively transcribed

The high steady state expression of K562 Alus might be determined transcriptionally or post-transcriptionally. The stability of RNA was observed in cells that had been treated with actinomycin to block transcription (19; Fig. 3A). 7SL RNA has a half-life of ∼10 h (Fig. 3A). In contrast, the abundance of flAlu RNA decays to 33% of its initial value within 30 min of blocking transcription in K562 cells (Fig. 3A). The decay of flAlu RNA in K562 and transfected 293 cells superimpose, indicating that flAlu RNA is not unusually stable in K562 cells (19; Fig. 3A).

Figure 3.

Stability and transcription of flAlu RNA in K562 cells. (A) K562 cells were exposed to actinomycin for the number of hours indicated and RNA was analyzed by primer extension for flAlu RNA and 7SL RNA. The intensity of these products was determined by phosphorimage analysis and is plotted in arbitrary units for the two RNAs. To facilitate direct comparison, the decay kinetics of flAlu RNA transiently expressed from clone XA in 293 cells (19) are also shown. (B) Nuclei from HeLa and K562 cells were transcribed in vitro with radiolabeled precursors and hybridized to equal molar quantities of DNA clones immobilized on filters. Filters corresponding to experiment 1 are shown. In calculating the relative intensities of these signals, pUC is used as a background correction (0 intensity) and the hybridization to Alu RNA (clone XAT) is taken as 1.

Nuclear run on assays compare the relative rates of Alu transcription in HeLa and K562 cells (Fig. 3B). Depending upon whether this comparison is made with 5S RNA or 18S rRNA, Alus are at least 4- or 6-fold more actively transcribed in K562 than in HeLa cells (Fig. 3B).

Methylation-mediated repression persists after stress

The transcriptional activities of methylated and unmethylated Alu constructs were compared in transient transfected K562, HeLa and 293 cells (Fig. 4 and data not shown). Transfected templates encode a homogenous length flAlu transcript which can be identified by northern analysis (Fig. 4A, lanes 1 and 2; 19). Additionally, a lower molecular weight product, scAlu RNA, results from processing flAlu RNA (19; Fig. 4). Endogenous 7SL RNA controls for RNA loading (Fig. 4). As a control for transfection efficiency, Southern analysis shows that DNA uptake was the same for the methylated and unmethylated templates and that both templates retained their methylation status following transfection (Fig. 4A and data not shown).

Figure 4.

Northern analysis of Alu RNA in transfected cells. K562 cells were transfected with methylated or unmethylated clone XAT and treated as described below. Using pUC DNA, the total amount of DNA transfected is adjusted to be 8 µg in all experiments. The lengths of the bands as assayed by northern hybridization correspond to those expected for 7SL RNA, flAlu RNA and scAlu RNA (12). The maximum variation of endogenous 7SL RNA in any series is 22%. ScAlu RNA bands are shown only for completeness and are not studied here. (A) Cells were transfected with either pUC alone (lane 1) or clone XAT (1, 5 and 8 µg in lanes 2, 3 and 4, respectively) or methylated clone XAT (1, 5 and 8 µg in lanes 5, 6 and 7, respectively). Southern analysis for clone XAT shows that methylated and unmethylated constructs have similar transfection efficiencies. (B) Cells were transfected with 5 µg of either XAT (lanes 1–4) or methylated XAT (lanes 5–8). These were used as untreated controls ‘C’ (lanes 1 and 5), or heat shocked and allowed to recover for 0, 1 or 4 h, as indicated at the top of the gel. (C) Cells were transfected with 5 µg of either XAT (lanes 1–4) or methylated XAT (lanes 5–8). These were used as untreated controls ‘C’ (lanes 1 and 5), or exposed to cycloheximide for either 0.75, 1.5 or 4.5 h, as indicated at the top of the gel.

Expression from unmethylated templates is 9–20-fold greater than from methylated templates in K562 cells (Fig. 4A). Methylation also inhibits expression in HeLa and 293 cells (not shown). The repressors which act upon methylated DNA are often limiting (29,30,34,51). In this case, the inhibition observed in these experiments would underestimate the actual repression that methylation causes and, indeed, the observed level of inhibition appears to be greater at lower template concentrations (Fig. 4A).

Presumably, Alu hypomethylation in K562 cells contributes to this higher transcriptional rate. An absence of methylation-mediated repression might further explain the relative insensitivity of K562 Alus to stresses. Conceivably, stress relieves the methylation-mediated repression which usually inhibits Alu transcription. This putative derepression would be transcriptionally regulated since neither short-term stresses nor longer-term viral infection have any observable effect upon methylation per se (39). In this model, stress would increase expression from methylated Alus (e.g. endogenous HeLa Alus) but would have less effect on the expression of unmethylated Alus (e.g. endogenous K562 Alus). We therefore compared the effects of cell insults on Alu expression from methylated and unmethylated templates.

Heat shock causes a slight increase (∼2–3-fold) in the expression of both unmethylated (Fig. 4B, lanes 1–4) and methylated templates in K562 cells (Fig. 4B, lanes 5–8). This modest increase is similar to the effect of heat shock on the expression of Alu RNA in K562 cells (Fig. 2). More importantly, expression from the methylated templates does not approach that of the unmethylated templates following heat shock (Fig. 4B). Similarly, cycloheximide initially causes a slight increase in expression from both methylated and unmethylated Alu templates in transiently-transfected K562 cells (Fig. 4C). Upon longer exposures to this drug, the level of Alu RNA may even decrease (8,17,18; Fig. 4C). Again, the level of expression from the methylated template is always far less than that of the unmethylated template (Fig. 4C). Similar results are obtained in transfected HeLa cells (data not shown).

An untested assumption underlying these experiments is that the methylated transfected templates faithfully report the regulation and stress response of endogenous templates. Although endogenous and exogenous templates might behave differently, the present results show that cell stress does not relieve the methylation-mediated repression that acts upon transfected templates. While not necessarily conclusive, this negative result encourages us to consider alternative pathways through which stress might increase Alu transcription.

Alu accessibility in K562 cell chromatin

We compared the chromatin structures in HeLa and K562 cells by digestion with HaeIII, HinfI and their combination (Fig. 5). HaeIII cleaves positions 1 and 45 and HinfI position 272 within the consensus Alu sequence (25,49). Each of these two restriction sites is ubiquitously distributed throughout the genome so that these enzymes also cleave non-Alu sequences. Ethidium staining detects total DNA (Fig. 5A), whereas Southern blot analysis detects Alu repeats using conditions that select for members of the Ya5 Alu subfamily which have mostly intact Hae and Hinf sites (Fig. 5B).

Figure 5.

Southern analysis of Alu chromatin accessibility. Nuclei from HeLa or K562 cells were undigested (lanes 1 and 5) or digested with either HaeIII (lanes 2 and 6), HinfI (lanes 3 and 7) or HaeIII and HinfI combined (lanes 4 and 8). Following DNA isolation purification, these samples were separated by agarose gel electrophoresis and visualized by either (A) ethidium bromide staining or (B) Southern hybridization for Alu DNA. M refers to a λPst marker and ‘Alu’ shows the position of the consensus 226 bp Hae–Hinf Alu restriction fragment.

Digestion with Hae or Hinf or their combination (Fig. 5A and B), or with micrococcal nuclease (data not shown) produces nucleosomal ladders for both HeLa and K562 chromatin. Mononucleosome bands in these restriction digests do not hybridize to the Alu probe (Fig. 5B). The lack of hybridization to mononucleosmal DNA is an expected consequence of the positions of both the hybridization probe and restriction sites within the Alu consensus sequence. Except for this difference, and a 226 bp Alu restriction fragment discussed below, the nucleosomal banding patterns observed by ethidium staining and Alu hybridization superimpose (Fig. 5A and B). The ladders in K562 and HeLa chromatin are noticeably shifted indicating a global difference in the chromatin structure within these two cell types (Fig. 5A and B). The estimated nucleosomal repeat unit lengths are 167 bp in HeLa and 179 bp in K562 cells with the difference rather than the absolute lengths being significant (Fig. 5). This difference is not peculiar to these restriction sites and is also observed in micrococcal nuclease digestions of HeLa and K562 chromatin (not shown). The reason for this gross difference in nucleosome packing is unknown.

Double digestion of K562 chromatin with Hae and Hinf liberates a consensus 226 bp Alu fragment which is barely detectable in HeLa chromatin (Fig. 5B, lanes 4 and 8). This band length is larger than the standard spacing of a mononucleosome but less than that of a dinucleosome, suggesting an altered chromatin conformation. Since restriction enzymes accessed both sites, either the Hae site or the Hinf site or both must be positioned at a gap within chromatin structure. We refer to these Alus as being accessible to restriction cleavage with the understanding that the Alu sequence contained within these structures need not be nucleosome free (46). Depending upon the extent of hybridization, a weaker 271 bp band is observed (Fig. 6 and data not shown). This longer band results from cleavage at the consensus HaeIII site, located at position 1 (25). This 271 bp fragment is also too short to be arranged within the standard dinucleosome spacing and results from accessible Alus.

Figure 6.

Effect of stresses on Alu chromatin. Nuclei from HeLa, 293 and K562 cells were digested with HaeIII and HinfI and the resulting DNAs were analyzed by Southern hybridization using an Alu probe. ‘C’ refers to untreated controls in the experiments described below. (A) Cells were infected with adenovirus ‘Adv’. The samples assayed here are identical to those assayed for Alu transcription in Figure 2A. (B) The blot used in (A) was rehybridized to an α-satellite probe. (C) Cells were treated with cycloheximide for 1 h (CX) or recovered for 30 min from heat shock (HS).

Stresses increase Alu chromatin accessibility

As assayed by in vitro transcription, adenovirus infection activates Alu transcription in HeLa cells by unmasking Alu chromatin structure (39). Similarly, adenovirus infection causes a dramatic increase in Alu chromatin accessibility in HeLa cells (Fig. 6A). There is also a very slight increase in nucleosome spacing following infection of HeLa cells, suggesting that this change in chromatin is not restricted to Alus but is more general (Fig. 6A; see internal marker below). In contrast, infection has little effect upon either Alu transcription (Fig. 2A) or chromatin accessibility in K562 cells (Fig. 6A).

HeLa and K562 cells differ in both Alu transcription (Fig. 2) and their nucleosome spacing (Fig. 5) either of which might be responsible for their difference in Alu chromatin accessibility. To distinguish between these alternatives, we investigated the accessibility of α-satellite DNA, a sequence which is not known to have any promoter activity (Fig. 6B). Human α-satellite is primarily organized as a tandem 340 bp repeat unit that is a dimer of divergent 170 bp monomer units (52). The principal product of cleaving α-satellite at either its consensus HaeIII or HinfI sites is 340 bp fragment whereas the product produced by cleaving both sites results is 216 bp (52; data not shown). Additionally, some 170 bp monomers are produced by cleavage with either enzyme (52). The chromatin structure of monkey α-satellite has been well characterized and while the 340 bp repeat is readily excised as a dinucleosomal restriction fragment (48,53), nucleosome periodicity should block double digestion with Hae and Hinf making the 216 bp fragment inaccessible. The similarity between this 216 bp Hae–Hinf fragment and the 226 bp fragment used to monitor Alu accessibility makes it an especially appropriate control. A ladder of α-satellite bands is released by Hae–Hinf digestion of chromatin (Fig. 6B). In contrast to the less precise spacing of nucleosome repeats, the positions and spacing of these bands corresponds to precisely defined α-satellite restriction fragments. Thus, some of these bands are resolvable as doublets and even triplets as a result of the different spacings of consensus restriction sites (Fig. 6B). The sharpness of these bands provides an internal marker for more subtle changes in nucleosome length that seems to occur upon adenovirus infection of HeLa cells (Fig. 6A and B; see above).

Of particular note, Hae–Hinf digestion releases a 216 bp ‘subdinuclosome’ α-satellite band from K562 cells and adenovirus infected HeLa cells but not from control HeLa cells (Fig. 6B). Like Alus, the chromatin accessibility of α-satellite is also greater in K562 and adenovirus-infected HeLa cells than in control HeLa cells (Fig. 6B). The opening in chromatin structure that adenovirus infection causes is not Alu-specific but even includes transcriptionally-inactive sequences.

Since chromatin accessibility increases independently of transcription, these results raise the question of whether transcription can increase independently of Alu chromatin accessibility. Heat shock and cycloheximide treatment each increase Alu chromatin accessibility in K562, HeLa and 293 cells (Fig. 6C). Compared to the effects of adenovirus infection on HeLa chromatin (Fig. 6A), some of these increases in Alu chromatin accessibility, e.g. 2-fold, are modest (Fig. 6C). However, adenovirus infection is an especially potent activator of Alu transcription in HeLa cells (7; Fig. 2A). These results indicate a tight correlation between Alu transcription and chromatin accessibility.

As estimated by phosphorimage analysis, approximately 0.1% of the Alus in control HeLa and 293 cells are in this accessible fraction as compared to about 2% of the Alus in untreated K562 cells (Fig. 6). For comparison, HeLa chromatin reduces Alu transcription in vitro by 100-fold relative to naked DNA (39). Results from two entirely different assays for the effect of chromatin on Alu accessibility agree within a factor of 10.

DISCUSSION

As expected, methylation inhibits transcription of transfected Alu templates in all cell lines tested, namely HeLa, 293 and K562. These observations complement an extensive literature showing that methylation inhibits Pol III transcription (2,29,30,34,51). Alus are usually hypermethylated, accounting for a major fraction of the total 5-meC content in human DNA (25–28). The obvious implication is that methylation generally represses Alu transcription. Since methylation inhibits transcription of transfected Alu templates in K562 cells, we infer that the extraordinary hypomethylation of Alus in these cells relaxes their usual repression and therefore partially accounts for their abnormally high level of Alu transcription.

While unmethylated Alu templates are efficiently expressed in transfected 293 cells, the same constructs are barely expressed in HeLa cells (19). Thus, demethylation is not sufficient for Alu transcription which must also depend upon cell-specific factors. Conversely, the relatively low expression of endogenous Alus in 293 cells (19) might be partially due to their hypermethylation (Fig. 1).

Heat shock, cycloheximide and adenovirus infection each increase Alu transcription in 293 and HeLa cells (7,8) but have relatively little effect upon K562 cells. These differences raised the possibility that stress might activate Alu transcription by relieving methylation-mediated repression. In entertaining this possibility, we note that Alu methylation patterns per se do not noticeably change following either viral infection or the far shorter time period associated with heat shock (39), so that this putative derepression would require a change in trans-acting factors. However, we observe no case in which either heat shock or cycloheximide relieve methylation-mediated repression in transfected cells. Although we remain open to the possibility that methylation affects transfected and endogenous templates differently, the methylation-mediated repression that is detected in these assays is stress insensitive. The separation of these effects implies that stress-induced activation and methylation-mediated repression work through separate activities.

Functional assays indicate that adenovirus infection and SV40 transformation stimulate SINE transcription by unmasking their chromatin structure making them accessible for in vitro transcription (12,39). Confirming those results, adenovirus infection similarly unmasks a subset of Alus in HeLa chromatin for restriction cleavage. While not directly tested, we think it apparent and, in this discussion, assume that this chromatin subset includes those Alus which are actively transcribed. Supporting this interpretation, Alus are also accessible for restriction cleavage in K562 cell chromatin and their accessibility increases following heat shock or cycloheximide treatment of HeLa and 293 cells. Together these results show a tight correlation between Alu transcription and chromatin accessibility. Previous studies show that chromatin represses Alu transcription in vitro (37,38).

This increased accessibility might either be required for or result from transcription and, furthermore, might reflect a change in either chromatin structure or transcription factor binding. The following reasoning suggests that this increase is due to an opening of chromatin structure that occurs prior to transcription: the increased chromatin accessibility of α-satellite DNA following viral infection implies that this change occurs independently of, and therefore presumably prior to, transcription and, further, that this change does not require factors that are specific for Pol III transcription. In addition, a mechanism by which viral infection specifically opens chromatin for Pol III-directed transcription has been established.

Viral infection increases the activity of TFIIIC2 which can derepress the effects of chromatin on Pol III-directed transcription (12,14,41,42,54 reviewed in 16). Activated TFIIIC relieves chromatin-mediated repression of Pol III-directed transcription through HAT activities, which reside on several subunits within TFIIIC2a (41,42,44,55). HAT activity might increase chromatin accessibility thereby derepressing Alu transcription in infected cells. Again according to this model, opening chromatin is a prerequisite for, rather than a result of, Alu transcription.

While the effect of viral infection on TFIIIC HAT activities suggests a plausible mechanism by which it might stimulate Alu transcription, viral infection of HeLa cells causes more global changes in chromatin as revealed by both the increased accessibility of α-satellite DNA and a slight increase in nucleosome spacing. We therefore conclude that other activities, not dedicated to Pol III transcription, are involved in these changes. Such factors might, nonetheless, stimulate Alu transcription by altering chromatin. As one possibility, histone H1 influences the length of linker DNA between nuclesomes, depleting H1 opens chromatin to nuclease digestion and H1 represses Pol III transcription of xenopus oocyte 5S genes and mouse B2 repeats (12,56–59). However, despite these attractive connections, depleting H1 from HeLa chromatin has little effect upon Alu promoter activity in vitro (37,38).

There are similarities and differences between Alu transcription in K562 and adenovirus-infected cells. As in virally infected cells, the higher level of Alu transcription in K562 cells is accompanied by a more open chromatin structure. The activity of TFIIIC is increased by viral infection as contrasted to the basal level of this factor in K562 cells. While TFIIIC activity in K562 cells remains to be determined, Alu hypomethylation in these cells would, as previously discussed, relieve the repression that methylation normally imposes thereby relaxing the requirements for Alu transcription.

Heat shock and cycloheximide also cause an increase in Alu chromatin accessibility. Based upon previous discussion, we infer that heat shock and cycloheximde open Alu chromatin as a prerequisite for, rather than a consequence of, transcription. Alu transcription responds rapidly, within 30 min, to either heat shock or cycloheximide (8) requiring similarly fast stress-induced changes in Alu chromatin structure. The observed changes in Alu chromatin structure do occur within 1 h of stressing cells.

Dynamic changes in chromatin evidently have a role in the complex regulation of this extremely abundant ‘multigene’ family. As emphasized in previous considerations of Alu transcription, the expression of a multigene family consisting of nearly one million members will be determined in many different ways (1). In addition to possible Alu-specific binding protein and more general regulators, such as H1 histone (12,37–40), individual Alu elements are also uniquely regulated through the cis action of their immediate flanking sequences and the activity of the chromosomal domain in which they reside (1,20–23,50). However, stress activates SINE transcription by remodeling chromatin to recruit additional templates.

DDBJ/EMBL/GenBank accession nos AF205194–AF205217