INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

The human -globin cluster, which extends over 70 kb of chromosome 11, contains the five -globin genes (, ^G, ^A, , and ) arranged in the same orientation and in the order of their expression during development. These genes are under the influence of a number of erythroid-specific transcription factors (TFs) including GATA-1 and its cofactor FOG-1 (30), NF-E2, and EKLF (reviewed in reference 21).

Several kilobases upstream of the -globin gene are at least five DNase I-hypersensitive sites (HS1 to HS5) which constitute the locus control region (LCR). These contain binding sites for GATA-1 and NF-E2 as well as CACC sites that bind a variety of general TFs including Sp1 (21). In transgenic mice the LCR is necessary for position-independent, copy-number-dependent regulated expression of the globin genes (13). Although HS2 has classical enhancer activity (27), recent experiments have shown that the LCR as a whole is orientation dependent and therefore does not fulfil the classical definition of an enhancer (28). It has been suggested that the LCR is responsible for the open chromatin structure of the -globin locus. Thus, in the naturally occurring Hispanic chromosome, from which 35 kb including most of the LCR is deleted, the entire cluster is DNase I resistant and the globin genes are transcriptionally inactive (11). However, recent discrete targeted deletions of both human and mouse LCRs in their natural contexts suggest that sequences outside the region covered by the five HSs may also play a role in the generation of a permissive chromatin structure throughout the cluster (10, 22).

Our previous studies of transcription in the -globin cluster showed that in addition to transcription of the five genic regions, there is also transcription in the regions between the globin genes and across the LCR (3). Transcription of the intergenic regions of the -globin cluster occurs on the same strand as that of the globin genes and is specific to erythroid cells. Further work on the phenomenon of intergenic transcription by Gribnau et al. (12) has suggested that the level of intergenic transcription can be correlated with the DNase I sensitivity in the -globin cluster. We have now extended our studies of intergenic transcription in the human -globin cluster. We have mapped the most 5' extent of LCR transcription into the endogenous locus to a retroviral element (ERV9) 5' to HS5. In contrast, intergenic transcripts downstream of two of the globin genes ( and ^A) do not correspond to any definable promoter and no precise transcription start site could be identified. We further show that intergenic transcription occurs across the entire human globin cluster in the adult erythroid cells of a transgenic mouse. Although intergenic transcription of the -globin cluster is specific to erythroid cells, it can be induced in nonerythroid cells by the expression of transiently transfected globin genes, a process known as transinduction. Even though the mechanism of transinduction is unknown, the transfected plasmids are seen to colocalize with the genomic copies of the cluster, and it is thought that this in some way releases the intergenic regions for transcription (3). We now describe two other means of inducing intergenic transcription in a nonerythroid environment, by the introduction of exogenous TFs and by the histone deacetylase inhibitor trichostatin A.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Plasmids and probes. Probes were generated from GenBank sequence accession no. AF064190 as follows: A (positions 1 to 512), B (512 to 1122), D (1941 to 2645), E (2645 to 3222), F (3222 to 3971), G (3971 to 4819), L17 (4818 to 5286), L18 (5286 to 5819), L19 (5819 to 6192), and L20 (6192 to 6509). Other globin probes were as previously described (3). Control probes for M13 (background), VA, and HIS and 5S (adenovirus VAI gene, mouse histone H4, and Xenopus laevis 5S DNA, respectively) have also been described (see references listed in reference 3).

Tissue culture. Subconfluent HeLa cells were transfected using Effectene (Qiagen) with 1 µg of VA plasmid, 1.5 µg of Tat plasmid and 5 µg of the globin and TF plasmids 24 h prior to nuclear run-on analysis (NRO). Hemin (40 µM) and trichostatin A (TSA) (3 µM) inductions were for 24 and 18 h prior to NRO, respectively.

NRO. Nuclear isolation and run-on analyses of cultured cells were performed as previously described (3). The line 72 transgenic mouse containing the human -globin cluster has been described (26). Mice were made anemic by treatment with 0.04 mg of acetylphenylhydrazine per g of body weight. An 80 to 95% pure population of erythroblasts was generated on the sixth day of treatment by gently scraping the surface of the excised spleen. These cells were washed in Tris-buffered saline (pH 7), and nuclei were prepared for NRO as for tissue culture cells.

RNA extraction and analysis. Cytoplasmic RNA was isolated 48 h after transfection as previously described (9). A BanII-AccI fragment encompassing the putative ERV9 promoter and -globin exon 1 in pF+ was cloned into pGEM4 and transcribed into an antisense riboprobe. Five hundred counts per second of riboprobe was hybridized to 20 µg of cytoplasmic RNA in 80% formamide-40 mM piperazine-N,N'-bis(2-ethanesulfonic acid) (PIPES) (pH 6.4)-1 mM EDTA-400 mM NaCl at 56°C overnight. RNase protection was as described previously (24).

RT-PCR analysis. One microgram of cytoplasmic RNA was reverse transcribed using primer G_RT (ATCTCCACATTATTTTAGTC) with Superscript II (Gibco BRL) according to the manufacturer's instructions. cDNA was amplified using Pfu Turbo (Stratagene) and the following primers spanning the ERV9 LTR: G_3' (TTAATTGCAAGGGCTGTTGGAAA), F_DOWN (TCGGGTCCCCTTCCACACTGT), F_DS (CTCTCTACCAATCAGCAGGATGT), F_U3 (ACCTTTGTGTGGACACTCTG), and F₆₀₀ (TTGCTGGGGATGCGAAGAAC).

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

The 5' extent of LCR transcription. Previous studies of the human -globin cluster showed that all regions of the LCR up to and including HS5 are transcribed (3). Using a variety of techniques, we now map the 5' extent of transcription beyond the LCR in the human erythroleukemic cell line K562, which endogenously expresses both - and -globin genes.

View larger version (38K): [in this window] [in a new window]   FIG. 1.   Transcription of the 5' region of the -globin LCR. (A) NRO analysis of hemin-induced K562 cells, in the absence and presence of -amanitin. The positions of single-stranded M13 probes relative to the ERV9 retroviral LTR and HS5 are shown. Probes for M13 background, histone H4 (HIS), and 5S RNA are also shown. (B) Analysis of endogenous LCR transcripts by RT-PCR. Primer positions are shown relative to the features of the ERV9 LTR. Red arrow, putative transcription start site; colored boxes, U3 tandem repeats. RT-PCR was carried out on cytoplasmic RNA extracted from K562 cells, and in each case water, RT-negative, and DNA controls were included. Green, LCR; red, ERV9 LTR.

View larger version (27K): [in this window] [in a new window]   FIG. 2.   Reporter constructs, whose structure is shown in the upper panel, were transiently transfected into HeLa cells and analyzed by RNase protection (A) and primer extension (B). The transcription start sites identified by primer extension are shown below panels A and B along with the positions of the 5' RACE products identified from endogenous LCR expression in hemin-induced K562 cells.

Definition of other intergenic transcripts. We have also attempted to assign start sites and promoter activity to the intergenic transcripts downstream of the human  and ^A genes. NRO suggests that intergenic transcription begins approximately 3.5 kb downstream of both  and ^A (3) (Fig. 4). Using the luciferase reporter gene system we found no evidence for promoter activity in either of these regions, despite testing a variety of fragments in the presence of an enhancer and in both erythroid and nonerythroid cells. We also attempted to detect 5' ends from endogenous intergenic transcripts in K562 cells using 5' RACE; however, we were able to detect only a low-level population of random ends (data not shown). Since some of these apparent 5' ends lie upstream of the region suggested by NRO, it is not clear whether they represented random initiation events. It is possible that in embryonic and fetal erythroid cells such as K562 a small amount of read-through from the expressed  and ^A genes occurs. Random premature termination of RT of such transcripts could then produce the population of transcripts we detected.

Intergenic transcription in a transgenic -globin cluster. We have also detected intergenic transcription in mice carrying the human -globin cluster as a transgene. These data therefore provide the first description of intergenic transcripts detected by NRO reading across the human -globin gene cluster in an authentic erythroid tissue. Previously, intergenic transcripts were analyzed in erythroid tissue only by in situ hybridization.

View larger version (42K): [in this window] [in a new window]   FIG. 3.   NRO analysis of intergenic transcription in a human -globin cluster transgene in the mouse. Probe positions are indicated (LCR probes are in green, genic probes are in blue, and intergenic probes are in black). The NRO profile of the transgenic mouse line 72 was compared to that of a nontransgenic mouse and with that of the human erythroleukemic cell line K562. Signals were quantitated by phosphorimager and corrected for U content and with respect to HIS and are shown in the lower panels. We have not looked for antisense LCR transcription in this mouse.

Induction of LCR transcription in nonerythroid cells. We have previously described one way of inducing intergenic transcription in the -globin cluster in nonerythroid cell lines, namely by transinduction. Transiently transfected plasmids expressing sequences within the -globin cluster colocalize with the chromosomal copy of the cluster and activate intergenic transcription in an as yet undetermined way (3). All regions of the -globin cluster tested (genic and intergenic) are able to transinduce intergenic transcription, and this effect is independent of the orientation of the expression plasmid (Fig. 4). This is entirely consistent with our previous suggestions that transinduction is mediated at the DNA level by the colocalization of the actively transcribed plasmid and the endogenous -globin locus (3).

View larger version (58K): [in this window] [in a new window]   FIG. 4.   Induction of transcription across the -globin cluster in nonerythroid cells. HeLa cells were transfected with a plasmid expressing the antisense -globin gene or with clones expressing the tethered NF-E2 heterodimer, GATA-1, FOG-1, or a GAL4/OCT-1 fusion protein plus the VA1 gene as a transfection control. (A) Cells were analyzed by NRO, and the resulting RNA was hybridized to a panel of probes taken from the LCR (green), intergenic (black), and genic (blue) regions of the entire cluster. (B) A more extended series of probes showing the beginning and end of sense strand transcription across the LCR is shown for HeLa cells expressing GATA-1, transinduced or induced with TSA, as well as for uninduced HeLa cells. Note that the three repetitive probes (E, F, and G) also give signals in the uninduced cells, presumably from copies elsewhere in the genome.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

These studies extend our knowledge of the nature and significance of nongenic transcripts derived from the human -globin gene cluster. We have used several approaches to determine where transcription of the -globin LCR begins in the human erythroleukemic cell line K562. These experiments reveal that there is no sense strand transcription upstream of an ERV9 endogenous retroviral element 1.1 kb upstream of HS5. This isolated LTR includes the retroviral promoter, which in transient assays was active in both erythroid and nonerythroid cells. However, although our analysis of endogenous RNA confirms that the ERV9 promoter is active and contributes to transcription of the LCR, transcripts that may initiate within the U3 repeats of the LTR were also detected. Our results extend the preliminary transcription analysis which accompanied a detailed sequence analysis of the region by Long et al. (18). However, these authors failed to detect transcripts initiating within the U3 region, and in this respect our analysis has been more refined. In view of the more extended region of DNase I sensitivity surrounding the -globin gene cluster, we have also looked for transcription on the antisense strand, towards the odorant receptor genes embedded within the same chromatin domain (5). A small amount of antisense transcription was seen over the repetitive probes (E through G) but did not extend into unique probes A through D or 17 through 20 (data not shown) and was likely to derive from ERV9 elements elsewhere in the genome. However, two additional probes which lie close to HOR5' 1 (GenBank accession no. AF137396, positions 29195 to 29737 and 30600 to 31024) did show erythroid-specific antisense transcription, which will be the subject of further studies.

Long et al. also suggest that the ERV9 LTR promoter has inherently erythroid properties. Although it is difficult to compare transfection experiments in different cell lines, our transient reporter gene analysis shows that ERV9 can be active in a nonerythroid environment at a level approaching its activity in erythroid cells. Furthermore, transinduction of HeLa cells shows that transcription from the endogenous -globin ERV9 LTR can be activated to quite high levels (comparable to those seen in K562 cells) in the absence of erythroid TFs, again suggesting that the promoters driving intergenic transcription are not inherently erythroid specific. By increasing the global level of histone acetylation using a histone deacetylase inhibitor (TSA), we were again able to induce intergenic transcription, which suggests that in nonerythroid cells the block to transcription is at the level of chromatin structure. A number of erythroid-specific TFs (NF-E2, GATA-1, and FOG-1) can also independently activate intergenic transcription. There are potential binding sites for GATA-1 within the U3 repeats of the ERV9 LTR (18) which may contribute to this activation. However, it is also possible that exogenously supplied TFs act through the LCR to open the chromatin which normally represses intergenic transcription in nonerythroid cells. We have not been able to activate transcription of the globin genes themselves, despite using a combination of the above induction methods. This may be because we have not accounted for all of the erythroid factors necessary for authentic globin gene expression, although it is also possible that the globin promoters are repressed in nonerythroid cells.

The finding that transcription of the LCR in part starts at the ERV9 LTR promoter is interesting because this region is not present in many transgenic mice which express the -globin genes normally. It is also absent from the endogenous mouse locus, the LCR of which is transcribed as shown by FISH analysis (3). It is therefore apparent that the ERV9 LTR is not required per se for LCR transcription. We have looked for transcription of the LCR in the transgenic mouse line 72, which has previously been described (13, 26) and which lacks the ERV9 LTR. Transcription of the human LCR was detected up to the most 5' probe present in the transgene (L17). We suggest that in this mouse the human LCR activates nearby endogenous mouse promoter(s) to serve as a transcription start site. However, it should be noted that our results do not rule out the possibility of additional transcription initiation events within the LCR, for example from HS2 (4, 31) and HS3 (17), as has been reported previously. It may be that initiation as far upstream as we observe is not essential for transcription of the LCR and that in some transgenic mice initiation at sites within the LCR provides the sole means by which it is transcribed. However, it is also possible that the absence of these upstream transcripts accounts for some of the minor position effects still seen in LCR-containing mice (2).

Our model of the ways in which transcription across the human -globin locus can be activated is shown in Fig. 5. In nonerythroid cells (Fig. 5a) the globin chromatin domain is normally closed so that no transcription occurs. In erythroid cells nearby promoters, which are not inherently erythroid specific, are activated, possibly through enhancer-type interactions with the LCR or else due to the partial opening of the -globin chromatin domain by the binding of erythroid TFs at the HSs (Fig. 5b). In the case of the endogenous human -globin gene cluster the most upstream promoter is in the ERV9 LTR, but in transgenic mice other promoters may be recruited. The globin genes are also activated by the combination of erythroid TFs binding to sites within their promoters as well as enhancer interactions with the LCR and the overall opening of the chromatin structure. Intergenic transcription can be induced in nonerythroid cells by at least three means: during transinduction by an unknown mechanism involving colocalization of the transfected plasmid and endogenous locus; by inhibition of histone deacetylases, resulting in a global opening of chromatin allowing access to these intergenic promoters (Fig. 5c); and by exogenously supplied erythroid TFs again either through enhancer-type interactions with the LCR or because of improved access due to the opening of the -globin chromatin domain (Fig. 5d). In each case the globin genes themselves are not transcribed, as the full complement of erythroid TFs are not present.

View larger version (15K): [in this window] [in a new window]   FIG. 5.   Transcription and chromatin structure in the globin locus in nonerythroid cells (a), in erythroid cells (b), following inhibition of HDACs with TSA (c), and following exogenous addition of erythroid TFs (d). The thickness of the solid black line represents the chromatin status: the thicker line indicates closed chromatin, while thinner lines indicate a more open structure. Green and blue arrows show intergenic and genic transcripts. Erythroid TFs (variously colored ovals) are shown bound to the 5 HSs of the LCR (green boxes) and to the promoter of the  gene (blue box). The solid and dashed black arrows indicate likely and possible, respectively, enhancer-type activations by the LCR.

While it has long been appreciated that the degree of chromatin condensation affects the ability of a region of DNA to be transcribed, it has recently become apparent that the opposite is also true and that transcription can affect the nature of chromatin. Thus, it has been demonstrated that the Saccharomyces cerevisiae elongator complex, which is integral to the transcribing polymerase II complex, has histone acetylase activity which may disrupt the chromatin structure facilitating rounds of transcription (34). Chromatin-remodeling factors have also been isolated as part of the polymerase II complex in both mammalian and yeast cells (7, 33). Such data suggest that intergenic transcription could maintain or extend the DNase I sensitivity initiated by the binding of TFs. Indeed, a recent report (17) shows that when erythroid TFs bind to the -globin HS2 and HS3 hypersensitivity is established over the core of the HSs in the absence of chromatin. Transcripts initiating within the core of both hypersensitive regions were also generated. A role for intergenic transcription in propagating DNase sensitivity in the -globin locus is clearly an attractive conclusion.

Support for this idea and for the importance of transcripts initiating upstream of HS5 comes from recent work by Reik et al. (22). A targeted deletion of HS2 to HS5 (2-5) in the human -globin LCR was generated so that the role of the LCR in globin gene expression could be assessed in its natural chromosomal location. Although globin gene expression was disrupted by this deletion, the cluster retained its erythroid-specific DNase I sensitivity. In contrast, the Hispanic thalassemia deletion causes the locus to become DNase I resistant in addition to affecting globin expression (11). Significantly, the Hispanic deletion includes the ERV9 LTR, whereas the 5' limit of the targeted deletion lies between HS5 and ERV9. We suggest that the DNase I sensitivity of the globin cluster is retained in the 2-5 mutation because transcription initiating within the ERV9 LTR can be activated by elements outside the deleted region; however, it has not yet been shown whether what remains of the LCR is transcribed following either of these deletions.

In a recent study of the human -globin cluster in transgenic mice, Gribnau et al. (12) suggest that the cluster can be divided into three differentially active subdomains: the embryonic and fetal subdomain, the fetal and adult subdomain, and an LCR subdomain active at all developmental stages. The production of intergenic transcripts in each case coincided with enhanced DNase I sensitivity and in the case of the and subdomains coincided with the activities of the genes themselves. Although the causal link between transcription and open chromatin suggested by these observations is attractive, our data are somewhat contradictory. We see equivalent levels of intergenic transcription in all three subdomains in adult (line 72 spleen) and embryonic (K562) cells and also during any of the induction methods that we have described. The discrepancy between our data and those of Gribnau et al. is likely to reflect the different sensitivities of the two techniques used, with NRO using several different probes providing a more quantitative measure of nascent transcription than does FISH.

Ultimate proof of a role for intergenic and LCR transcription in maintaining an open chromatin structure and through this mechanism affecting -globin gene expression must come from disrupting these transcripts in a whole-locus setting (either in transgenic mice or by targeted disruption of the endogenous locus in cell lines). While those experiments are currently in progress, the data presented here and recent work by others lead us to predict that nongenic transcription will have an important role in these processes. Furthermore, the demonstration that the intergenic regions of other gene clusters (e.g., IL-4/IL-13 [23]) are transcribed leads us to anticipate that intergenic transcription will be a feature of at least some mammalian gene loci.