INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

Recombinant retroviruses derived from murine leukemia viruses (MuLV) are widely used as vectors for gene transfer into a variety of cell types, in both research and clinical applications (36). However, retroviruses are highly susceptible to transcriptional silencing and position effects imparted by chromosomal sequences at their integration sites (4, 30, 36). While silencing of viral and transposed elements may protect the genome from insertional mutagenesis (4), it also acts to repress the expression of retrovirally transduced genes (5, 6, 13, 18, 28). Retrovirus expression is also greatly influenced by endogenous enhancers and heterochromatic regions near the integration site (16). These effects curtail the correct regulation and sustained expression of retrovirally transduced genes and thus represent a major obstacle to the therapeutic use of recombinant retroviruses (30, 36). Retroviral methylation is commonly associated with transcriptional silencing (15, 17, 18, 23). Methylation of retroviral vectors occurs in a number of tissues and is likewise associated with decreased transgene expression in vivo (4, 5, 30, 31). These observations and the recent finding that methyl-CpG-binding protein 2 binds to methylated promoters and recruits histone deacetylases which are able to repress transcription (21, 26) suggests that DNA methylation may play a primary role in the silencing of retroviruses. A better understanding of these mechanisms is needed to devise novel approaches to overcome retroviral vector silencing.

Insulators are DNA sequences that can function as directional blocking elements either by interfering with promoter-enhancer interactions when positioned in the intervening sequence or by reducing position effects imparted on transgenes when flanking the integrated transcription units (2, 7, 12). Insulator elements have been shown to reduce position effects in transgenic animals, particularly in Drosophila (9, 12) and to a lesser extent in vertebrates (7, 19, 20). The most characterized vertebrate element is the chicken hypersensitive site 4 (cHS4), an insulator sequence of the chicken -like globin gene cluster. It has been shown to prevent position effect variegation in transgenic Drosophila (7, 8), mice (39), and rabbits (35) and in the chicken erythroid cell line 6C2, where it also exerts an antagonist effect on transgene methylation (29).

Recently, a minimal core element of the insulator has been characterized in more detail and a putative binding protein, called CCCTC-binding factor, has been identified (3). In the human cell line K562, the minimal core insulator element has been shown to have enhancer-blocking activity when placed between the enhancer and the promoter of a reporter gene (3, 7). In one study using homologous recombination to analyze different constructs in the same chromosomal locus, the insulator reduced enhancer activity when placed on the distal flank of the enhancer relative to the promoter. This suggests that the insulator may also have silencing activity, at least at some chromosomal sites (37). Altogether these results suggest that additional experimentation is needed to better characterize insulator function and to assess whether insulator elements could be useful to improve transgene expression in gene therapy applications.

We therefore incorporated the cHS4 element upstream of the retroviral enhancer/promoter sequence of a recombinant MuLV. In contrast to other studies (3, 7, 8, 14, 23, 31, 37), transduction was performed at high efficiency by retroviral infection without any selection, which would unavoidably bias the analysis toward favorable integration sites. Thus, all integration sites were amenable to molecular analyses. We show that in murine erythroleukemia (MEL) cells, cHS4 increases the probability that randomly integrated proviruses will express. cHS4 dramatically decreases vector methylation, and we show that protection from methylation occurs in the absence of transcription from the long terminal repeat (LTR). In embryonic stem (ES) cells, however, retroviral vectors bearing the insulators do not express the marker gene and the LTR is completely methylated within 6 days after retroviral transduction. Surprisingly, cHS4 has little effect on positional variability of expression, indicating that it does not confer uniform position-independent expression from retroviral vectors, nor does it create DNase I-sensitive borders between the retroviral transcriptional unit and flanking chromatin at all integration sites. These results suggest that cHS4 may be useful but not sufficient to overcome silencing associated with methylation of recombinant retroviruses.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Vector construction. To construct the ISN vectors, the 1.2-kb cHS4 sequence (gift from G. Felsenfeld and M. Reitman) was digested from plasmid pJC5-4 (8) with XbaI, blunted, and cloned into the NheI-blunted site of the 3' LTR of MuLV strain SN (11), generating the vectors I1SN and I2SN ("1" corresponds to the insulator element cloned as shown in Fig. 1A; "2" corresponds to the opposite orientation). As a control, a fragment of comparable size taken from the glyceraldehyde-6-phosphate dehydrogenase (G6PD) cDNA (+873 to +1656) was cloned at the NheI site (JSN). DSN vector was generated by deleting the 3' U3 region from the PvuII-to-BssHII sites to remove the enhancer/promoter elements. To construct the I1DSN vectors, the 1.2-kb cHS4 sequence was digested from plasmid pJC5-4 with XbaI, blunted, and cloned into the NheI-blunted site of the 3' LTR of DSN. I1-SacI DNS was generated digesting the cHS4 element cloned in the SacI site from plasmid pJC5-4 and inserted in the SacI sites of the 3' LTR of DSN.

View larger version (50K): [in this window] [in a new window]   FIG. 1.   Recombinant retroviruses bearing cHS4 stably integrate with the intact insulator sequence. (A) Map of SN, I1SN, JSN, and DSN. To construct the ISN vectors, the 1.2-kb cHS4 sequence was cloned into the NheI (I1SN and I2SN) site of the 3' LTR of SN (11). As a control, a fragment of comparable size taken from the G6PD cDNA (+873 to +1656) was cloned at the NheI site (JSN). DSN vectors were deleted in the 3' U3 region to remove the enhancer/promoter elements (see Materials and Methods). Letters in italics indicate positions of restriction enzyme sites: N, NheI; B, BssHII; S, SacI; H, HindIII; Ba, BamHI. I1SN and JSN harbor the inserted sequences at the 3' NheI site of SN. SD, splice donor; SA, splice acceptor. (B) Southern blot analyses. Genomic DNA extracted from NIH 3T3 fibroblasts was digested with the enzyme indicated above each lane. The left blot was probed both with NTP and cHS4 sequences; the right blot was probed with NTP only. Sizes are indicated in kilobases. E.B., endogenous bands.

Virus production, NIH 3T3, MEL, and ES cell infection. Cell-free viral stocks were generated from gpg29 packaging cells as previously described (11) and then titrated by Southern blot analysis and concentrated as described elsewhere (11). NIH 3T3 and C88 MEL cells were transduced as described elsewhere (32). The ES cell line CJ7 was maintained on gelatin-coated tissue culture plates in Dulbecco modified Eagle medium supplemented with 15% ES serum, penicillin-streptomycin (50 U/ml), L-glutamine (2 mM), and 0.1 mM (final concentration) -mercaptoethanol, with the addition of recombinant leukemia inhibitory factor (0.1 ng/ml, final concentration) to preserve their undifferentiated state. C88 MEL cells were infected at a multiplicity of infection of 2 (to generate single-copy clones) or 2 to 10 (for studies in MEL cell populations [Fig. 5]) in the presence of Polybrene (Sigma, St. Louis, Mo.) at 4 µg/ml. The ES and NIH 3T3 cells were transduced in their own media for 16 h at multiplicities of infection of 3, 10, and 30 in the presence of Polybrene (Sigma) at 4 or 8 µg/ml.

PCR and Southern blot analyses. MEL cells were subcloned by limiting dilution in 96-multiwell dishes and screened for transduction using primers within the coding sequence of NTP (a mutated form of the human low-affinity nerve growth factor [11]). The NTP gene was amplified with the oligonucleotides NTP F1 (5'-CTTGGAGGTGCCAAGGAGGCATG-3') and NTP R4 (5'-CCAGCGTGTGCACTCGCGGA-3') for 40 cycles of 94°C for 1 min, 63°C for 1 min, and 72°C for 1 min.

NTP expression assays. Immunofluorescent detection of cell surface NTP was performed by fluorescence-activated cell sorting (FACS) analysis (FACScan; Becton Dickinson) using the anti-LNGFR monoclonal antibody 20.4 (American Type Culture Collection, Manassas, Va.) as previously described (11). Live cells were gated based on forward scatter and side scatter. Northern blot analyses were performed with 10 µg of total RNA and the NTP cDNA and -actin gene as probes.

DNase I assay. Nuclei were prepared as described elsewhere (33). Aliquots of 3 × 10⁸ nuclei were treated with 2.5 or with 0.5 µg of DNase I (Boehringer Mannheim) for fixed increments of time. Each reaction was performed in 0.5 ml of TMSD solution (MgCl₂, 1.5 mM; Tris-HCl [pH 7.5], 10 mM; dithiothreitol, 0.5 mM; phenylmethylsulfonyl fluoride, 1 mM; sucrose, 0.25 M; CaCl, 10⁴ M). The reactions were performed at 37°C and stopped after variable times by adding 130 µl of stop buffer (30 µl of 10% sodium dodecyl sulfate, 100 µl of 0.25 M EDTA with proteinase K [Boehringer Mannheim] at 0.1 mg/ml [final concentration]) at 56°C for 2 h. The DNAs digested with 2.5 µg of DNase I were subjected to phenol-chloroform extraction, ethanol precipitation, and NheI restriction digestion to analyze retroviral sequences (Fig. 6A). The DNAs digested with 0.5 µg of DNase I were subjected to phenol-chloroform extraction, ethanol precipitation, and BamHI restriction digestion to examine 5' flanking chromosomal sequences (Fig. 6B). Blotted digests were probed with the NTP and -actin cDNAs.

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

The cHS4 insulator sequence inserted in the 3' Mo-MuLV LTR is faithfully duplicated after retroviral integration. The cHS4 insulator was cloned into the LTR of SN, a replication-incompetent Moloney MuLV (Mo-MuLV) that encodes an inert cell surface marker termed NTP (11). The 1.2-kb genomic fragment (8) was introduced in the 3' U3 region, upstream of the retroviral enhancer/promoter, to create the two vectors I1SN (Fig. 1A) and I2SN (not shown). Thus, following reverse transcription and integration, a duplicated insulator flanks the retroviral transcription unit at both ends. As a control, we replaced the insulator with a portion of a human cDNA sequence, corresponding to the translated region from nucleotides +813 to +1656 of the human G6PD gene (JSN vector) (Fig. 1A) (24). To investigate the stability and level of expression of the NTP gene marker, we infected NIH 3T3 cells with recombinant virions pseudotyped with the vesicular stomatitis virus G glycoprotein (11). Southern blot analyses showed that retrovirally infected cells harbored stably integrated vectors with intact recombinant LTRs after reverse transcription and integration (Fig. 1B). In NIH 3T3 fibroblasts, which are highly permissive for retroviral infection and expression, SN, I1SN, and I2SN yielded identical NTP expression in populations harboring the same average vector copy number (data not shown). Insulator orientation did not interfere with stable integration of the vector and expression of the NTP gene; the I1SN vector (Fig. 1A) was selected for further studies.

Efficient gene transfer in MEL cells allows for the generation of a panel of single-copy integrants without any selection bias. MEL cells were infected with recombinant virions pseudotyped with the vesicular stomatitis virus G glycoprotein (Fig. 2A). Strong position effects have been previously reported for MEL cells (32, 34). To follow the expression and structure of the recombinant genomes integrated at different chromosomal positions, we generated a large panel of MEL cell clones bearing SN, I1SN, or JSN. We devised a strategy that avoids selection based on transgene expression to ensure that all integration sites could be examined, including sites where retroviral expression is silenced or very weak. Conventional drug selection would indeed eliminate the latter cells and bias any analysis toward the subset of integration sites that are permissive for a threshold expression level compatible with drug resistance. MEL cells were therefore subcloned by limiting dilution immediately after retroviral transduction and scored for vector integration by PCR analysis (Fig. 2C and D). More than 600 clones derived from three independent infections were analyzed by PCR. Positive clones were further expanded. Vector copy number was determined by Southern blot analysis using NcoI digests. NcoI recognizes only one site in the integrated retroviral sequence, generating a single band for each different genomic integration site (Fig. 2E). Single-copy and low-copy-number clones were retained for this study to permit the direct correlation of transgene expression with the corresponding vector integration site (Fig. 2F and G).

View larger version (47K): [in this window] [in a new window]   FIG. 2.   Analysis scheme. A large panel of MEL cell clones bearing SN, I1SN, or JSN was generated and analyzed. MEL cells were infected with recombinant virions pseudotyped with the vesicular stomatitis virus G glycoprotein (A) and analyzed by FACS (B). MEL cells were then subcloned by limiting dilution immediately after retroviral transduction (C) and scored for vector integration by PCR analysis (D). N.B., NTP band, amplified by PCR, indicating the presence of the integrated proviral sequence. (E) Vector copy number was determined by Southern blot analysis using NcoI digests. NcoI recognizes only one site in the integrated retroviral sequence, generating a single band for each different genomic integration site. The Southern blot analysis shows vector copy number for 11 I1SN MEL clones. All MEL clones derived from the third selection (III in Fig. 3A). The blot was probed with the NTP sequence. E.B., endogenous bands. (F and G) Randomly chosen clones were retained to permit the direct correlation of transgene expression with the corresponding vector integration site.

The cHS4 insulator increases the probability of retroviral expression at random chromosomal integration sites. We identified 23 single-copy clones transduced with the SN vector, 34 transduced with I1SN, and 11 transduced with JSN and measured NTP expression by serial FACS analyses from days 21 to 57 after retroviral infection. At day 21, we found that the vast majority of clones transduced with SN or JSN (27 of 34 [79%] altogether) failed to express detectable NTP by FACS (Fig. 3A). In contrast, 74% (25 of 34) of the clones transduced with I1SN expressed NTP. By day 57, 44% of the I1SN clones showed detectable NTP expression, in contrast to only 13% (3 of 23) of the SN clones (Fig. 3A) and none of the 11 JSN clones (data not shown). We therefore concluded that the presence of the duplicated cHS4 insulator increased the probability that a transduced cell would express the retrovirus-encoded transgene. In all clones, immunofluorescence of NTP expression showed narrow single peaks. Double peaks or broad peaks, which would have been suggestive of variegated transgene expression, were not observed (e.g., Fig. 3B). Measurements by FACS analysis were corroborated by Northern blot analysis. Total RNA, extracted from 11 I1SN, 3 SN, and 3 JSN clones 57 days postinfection (dpi), was analyzed with NTP. The nine clones that were negative by FACS analysis, SN-27, SN-42, SN-A12, JSN-5, JSN-17, JSN-18, I1SN-23, I1SN-43, and I1SN-60 (data for SN and I1SN in Fig. 3A; JSN data not shown) were also negative at the RNA level (Fig. 4). Thus the FACS analysis matched the RNA analysis except for one clone (I1SN-28) in which a very low level of NTP expression was detected by Northern blot (Fig. 4) but not FACS analysis.

View larger version (44K): [in this window] [in a new window]   FIG. 3.   NTP expression measured at the protein level by FACS analysis. NTP expression was determined in 57 clones transduced with a single copy retroviral vector 21 and 57 dpi. Levels of expression are expressed as the percentage of cells brighter than background autofluorescence to provide a measurement reflecting the average level of expression for each clone (NTP+ cells; y axis). (A) NTP expression in 23 single-copy clones transduced with the SN vector, 34 transduced with I1SN, and 11 transduced with JSN was measured by serial FACS analyses from days 21 to 57 after retroviral infection. At day 21, we found that the vast majority of clones transduced with SN or JSN (27 of 34 altogether) failed to express detectable NTP by FACS. In contrast, 74% (25 of 34) of the clones transduced with I1SN expressed NTP. By day 57, 44% of the I1SN clones showed detectable NTP expression, in contrast to only 13% (3 of 23) of the SN clones and none of the 11 JSN clones (data not shown). (B) NTP expression in four clones measured by FACS analysis at 57 dpi. In all clones, the immunofluorescent NTP signal showed narrow single peaks. Double peaks or broad peaks, which would have been suggestive of variegated transgene expression, were not observed at 21 or 57 dpi.

View larger version (42K): [in this window] [in a new window]   FIG. 4.   NTP expression measured by Northern blot analysis. Total RNA extracted from 11 I1SN (8 single-copy clones, 2 with two copies, and 1 with three copies), 3 SN single-copy clones, and 3 JSN single-copy clones at 57 dpi was probed with the NTP (top) and human -actin (bottom) sequences. U and S represent the unspliced and spliced transcripts, respectively. The control DNA in the Northern blot on the right corresponds to RNA from an I1SN clone loaded with the SN or JSN RNA on the same gel as control for the NTP probe. Nine clones that were negative by FACS analysis (SN-27, SN-42, SN-A12, JSN-5, JSN-17, JSN-18, I1SN-23, I1SN-43, and I1SN-60) were also negative at the RNA level.

Methylation analysis reveals that the cHS4 insulator element dramatically decreases the level of methylation in both the 5' and 3' LTRs. Proviral methylation was investigated in a representative subset of clones (Fig. 5A). Genomic DNA of 10 single-copy SN clones and 11 I1SN clones (8 with one copy and 3 with two or three vector copies [Fig. 2E]) were randomly chosen and analyzed 21 and 57 dpi. DNAs were digested with NheI and BssHII. NheI is a restriction enzyme that recognizes a sequence located in the U3 region of the LTR, while BssHII is a methylation-sensitive restriction enzyme specific for a single site located between the CAAT and TATA boxes in the LTR (Fig. 1A). The expected bands are 2.9 kb (BssHII site methylated in the 5' LTR) or 2.5 (BssHII site unmethylated). At day 21, seven of nine SN clones negative for NTP expression were heavily methylated in the 5' LTR (methylation index [MI] greater than 50%; see legend). By day 57, the MI increased to 90% or more. Clone SN-31 (positive for expression [Fig. 3A]) was significantly less methylated (MI of <50% at day 57). In the I1SN clones, 10 of 11 showed remarkably little methylation on day 21 in the 5' LTR. By day 57, three showed a higher MI, but the average level of methylation remained very low (23%, n = 11). The three NTP-negative (both by FACS and by Northern blotting) clones (I1SN-23, -43, and 60 [Fig. 3A]) were the only three I1SN clones to be heavily methylated.

View larger version (49K): [in this window] [in a new window]   FIG. 5.   Analysis of vector methylation in MEL cell clones. Proviral methylation was investigated in a representative subset of clones. Genomic DNA of 10 single-copy SN clones and 11 I1SN clones (8 single-copy and 3 with two or three vector copies [Fig. 2E]) were randomly selected and analyzed 21 and 57 dpi. (A) Southern blots showing SN or I1SN DNA clones digested with BssHII (B) and NheI (N). The sizes of the expected bands are indicated on the side: BssHII site methylated (M) in the 5' LTR (2.9 kb, top band) or BssHII site unmethylated (U) (2.5 kb, bottom band). (B) The same DNAs were digested with BssHII alone. The expected bands are 2.9 and 4.1 kb, respectively, for SN or I1SN when the two LTR BssHII sites remain unmethylated. In the single BssHII digest, clone I1SN-45 is not shown. All MEL clones were derived from the third selection (III in Fig. 3A). The MI for the 5' LTR in the BssHII site is calculated as the percentage of the unmethylated band over the sum of the unmethylated and methylated bands.

View this table: [in this window] [in a new window]   TABLE 2.   Wilcoxon rank sum test to compare percentage DNA methylation in NTP+ and NTP clones

The insulator is DNase I resistant in silenced integration sites. To assess chromatin structure at the proviral integration sites, DNase I sensitivity was measured in a series of clones, including SN-27, SN-A12, JSN-5, and all 11 I1SN clones. Nuclei from these clones were digested with the NheI after increasing amounts of time of DNase I treatment (1, 5, 13, and 17 min), and their DNA was probed with the NTP sequence. As shown in Fig. 6A, in all NTP-positive clones, the proviral band (NheI digest) disappeared in less than 1 min of treatment (clones I1SN-18, -28, -29, -45, -53, -54, -61, and -77). In contrast, clones not expressing NTP (SN-A12, SN-27, JSN-5, and I1SN-23, -43, and -60) all resisted DNase I digestion for 5 min or more. The clones with the highest methylation indices were the most DNase I resistant (e.g., I1SN-23 and JSN-5 [Fig. 6A]). Thus, we found a very strong correlation between NTP expression, promoter methylation, and DNase I sensitivity. The majority of integrated retroviral sequences flanked by cHS4 show an open chromatin structure sensitive to DNase I treatment.

View larger version (63K): [in this window] [in a new window]   FIG. 6.   DNase I sensitivity of retroviral and upstream flanking chromatin. (A) Nuclei from 11 I1SN clones, 2 SN clones, and 1 JSN clone were digested with NheI after increasing amounts of time of DNase I treatment (1, 5, 13, and 17 min), and the DNAs were probed with the NTP gene. NTP expression by FACS and percentage of methylation of the 5' LTR in BssHII are represented for each clone. (B) The three I1SN clones most DNase I resistant in the preceding experiment were also analyzed with BamHI after DNase I treatment for 1, 2, 4, and 8 min. BamHI recognizes only one site at the 3' end of the NTP gene in the proviral integrated sequence (Fig. 1A), thus generating a unique band for each genomic integration site. All the blots were probed with NTP and subsequently with the human -actin sequence (not shown) to assess the quantity of loaded DNA and the homogeneity of DNase I treatment within the clones. E.B., endogenous bands.

The insulator element does not prevent Mo-MuLV LTR methylation in ES cells. Gene expression from the Mo-MuLV virus is restricted in embryonal carcinoma and ES cells (6, 23, 40). Moreover, the loss of expression in ES cells is directly correlated to methylation of the LTR (31). We decided to analyze whether the modified LTR bearing the insulator element was able to prevent methylation of the LTR and rescue the vector from loss of expression in ES cells. We infected ES and NIH 3T3 cells with different concentrations of the SN or I1SN vector and monitored NTP expression by FACS 3 and 6 dpi. Moreover, 6 dpi, DNA was extracted and digested with NheI and BssHII to evaluate the efficiency of infection and methylation of the LTR in both NIH 3T3 and ES cells (Fig. 7). We observed, as expected, that NIH 3T3 cells showed high levels of expression of the NTP marker gene in all groups of infected cells (data not shown) and lacked LTR methylation. In contrast, the ES cells were negative by FACS (data not shown) and the LTRs were completely methylated, suggesting that the insulator element is unable to prevent the methylation mechanism from acting on the LTR in ES cells.

View larger version (45K): [in this window] [in a new window]   FIG. 7.   Analysis of vector methylation in 3T3 and ES cells. ES and NIH 3T3 cells were infected with different concentrations of the SN or I1SN vector. Genomic DNA was extracted 6 days later and digested with BssHII and NheI to evaluate the efficiency of infection and the methylation of the LTR. Multiplicity of infection (MOI) is represented for each lane below the Southern blots. The Southern blot was probed with the NTP sequence. The sizes (in kilobases) of the expected bands are indicated on the side (top band, methylated; bottom band, unmethylated).

The cHS4 insulator element prevents LTR methylation in the absence of transcription. As the LTR is poorly, if at all (5), expressed in ES cells, we next examined whether the activity of cHS4 depended on retroviral transcriptional activity. We disabled the LTR enhancer/promoter via an extensive deletion in the U3 region that removed the direct repeats of the enhancer and promoter up to the TATA box (DSN [Fig. 1A; Materials and Methods]).

View larger version (73K): [in this window] [in a new window]   FIG. 8.   Prevention of retroviral methylation does not require transcriptional activation. Shown is methylation analysis of MEL cell pools infected at the same multiplicity of infection with the SN, I1SN, DSN, and I1DSN vectors. In DSN and I1DSN vectors, the LTR enhancer/promoter is disabled via an extensive deletion in the U3 region that removes the direct repeats of the enhancer and the promoter up to the TATA box (Fig. 1A; Materials and Methods). Proviral methylation was monitored for 60 days after infection: data are shown for the MEL pools analyzed 60 dpi. Restriction enzymes are indicated as follows: N, NheI; B, BssHII; S, SacI; H, HindIII; Ba, BamHI. The sizes (in kilobases) of the expected bands are indicated on the side (top band; methylated; bottom band, unmethylated), and the MI is given below each lane. The MI for the 5' LTR in the BssHII site is calculated as the percentage of the unmethylated band over the sum of the unmethylated and methylated bands. NTP expression was undetectable by FACS and Northern blot analyses in DSN and IDSN MEL cells (data not shown).

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

To study the effect of the cHS4 insulator on retroviral expression, we introduced recombinant retroviruses in MEL and ES cells, in which strong position effects and retrovirus silencing are known to occur. In MEL cells bearing a single copy of the different recombinant retroviruses, we observed that the cHS4 insulator increases the probability of expression at random integration sites from 7 of 34 (or 21%) to 25 of 34 (or 74%) (Fig. 3A). The positive clones express protein and transgene mRNA at very different levels (Fig. 3B and 4), showing that the cHS4 did not result in uniform gene expression levels. In studies where position-independent expression was suggested (7, 8, 29), a selectable marker was used to generate cells carrying the insulator element. However, genetic selection biases the analysis toward a subset of integration sites that are permissive for a minimum threshold expression level compatible with drug resistance. Selection will therefore eliminate silent or very unfavorable integration sites and thus appear to reduce the variability of expression between clones. In our system, no selective pressure was applied to the clones, enabling us to enumerate and analyze all integration sites, including the most unfavorable. Furthermore, we could measure the level of transgene expression (Fig. 3A and 4) and relate it to the function of a single transcription unit through the identification of single-copy clones (Fig. 3E). We found that cHS4 increased the probability that the vector would express the transgene (Fig. 3A and 4). To that extent, the insulator reduced position effects. However, clones expressing NTP still varied greatly in their level of expression, suggesting that the cHS4 insulator could not alone create conditions for uniform expression, at least in retrovirally transduced MEL cells.

We found that the insulator element prevents and/or delays methylation of the LTR. Methylation of the BssHII site was strongly associated with lack of expression (Table 1) in both the SN and I1SN clones. These findings indicated that the chance that the SN vector would be silenced and methylated was position dependent, occurring at 8 of 10 sites. Incorporation of the insulator concomitantly reduced retroviral silencing and methylation at most but not all sites, as 3 of 15 were silenced nonetheless. These findings do not distinguish whether methylation precedes or follows vector silencing. When examining the chromatin structure in the 5' flanking region of the integrated retrovirus in the three silenced I1SN clones, we found that cHS4 did not mark a transition in DNase I- sensitivity between retroviral and chromosomal sequences (Fig. 6B). The strong association between methylation and DNase I- resistance does not allow us to identify which of the two chromatin alterations precedes the other in causing transcriptional silencing. However, our findings in MEL and ES cells are most consistent with a silencing mechanism primarily driven by proviral methylation because cHS4 has itself no transcriptional ability (7, 37) and prevented methylation independently of promoter activation (Fig. 8). This model is consistent with a mechanism whereby prevention of methylation preempts secondary chromatin condensation (21) and suggests that there are two major categories of retroviral integration sites. In this model, a recombinant retrovirus like SN escapes silencing only if it integrates at very favorable chromosomal sites (e.g., close to a CpG island or actively transcribed sequences), about a quarter of all sites in MEL cells (Fig. 3A). In the other sites, silencing will prevail. The cHS4 insulator renders the retroviral sequences less susceptible to methylation at a majority of integration sites (Fig. 5), except for a subset of most unfavorable sites, perhaps sites located in centromeric or subtelomeric regions (10, 27) where silencing mechanisms may be stronger or of a different nature.

The lack of transgene expression that we observed in ES cells was strongly associated with methylation of the LTR (Fig. 7). Loss of expression in ES cells has been previously correlated with methylation of the LTR (31), and modifications of the LTR sequence that allowed low-level expression in ES cells inversely correlated with methylation of the LTR (31). This suggests that a strong active silencing mechanism acts on the Mo-MuLV promoter in ES cells. Part of this repression may be explained by the binding of transcriptional repressors expressed in primitive embryonic cells (22, 40). There are several possible explanations to the apparent lack of an insulator effect in ES cells. It may be due to erythroid specificity of cHS4. cHS4 has indeed been mostly investigated in erythroid cell lines such as K562, C12, and MEL (references 3, 8, 29, and 37 and our data). However, there are reports suggesting activity that cHS4 is active in other lineages (35, 39). This does not exclude that cHS4 may have greater or additional activity in erythroid cells due to the possible lack of expression of proteins required to activate cHS4 in ES cells (3). Alternatively, methylation processes may either be qualitatively or quantitatively different in ES cells, precluding any effects of cHS4. Another possibility is that prevention of de novo methylation requires transcriptional activation of the LTR, which may not occur in ES cells. In this case, prevention of LTR methylation would be irrelevant to activate the LTR, although it has been shown to be essential to maintain active transcription (31). Thus, the effects of the insulator would be completely masked by the absence of transcription from the LTR. However, lack of transcription from the LTR did not lead to higher levels of methylation, as we showed in MEL cells transduced with vectors carrying intact or deleted promoters (Fig. 8). It is noteworthy that the only retroviral vectors that previously showed transgene expression in ES cells were isolated under drug selection (6, 23, 31). As our studies were performed without exerting selective pressure on the ES cells, we cannot exclude that a very small minority of cells expressed the marker gene and remained unmethylated. Interestingly, in MEL cells, the presence of cHS4 exerted its effect at a majority of integration sites, but not all of them, suggesting that either the propensity to methylate the integrated retroviral sequence or the activation of cHS4 activity varies in different chromosomal regions.

For gene therapy applications, it will be important to define the scope of cell types in which the insulator can increase the probability of transgene expression and/or prevent vector silencing. Importantly, because insulator activity does not require retroviral expression, it could be useful in gene therapy applications where transcriptional activation of the vector occurs only after target cell differentiation in vivo (30). Thus, the insulator may prove valuable in the transfer of tissue-specific vectors in hematopoietic stem cells, such as -globin gene vectors (32), which remain silent in the transduced stem cells and activated only after some of the differentiated progeny matures into proerythroblasts. It also remains to be investigated whether insulators favor position-independent expression if they are used in conjunction with appropriate transcriptional regulators or other determinants of chromatin structure (1, 32).