Positive and Negative Regulation of the Human Insulin Gene by Multiple Trans-acting Factors*

Tissue-specific expression of the human insulin gene is regulated by &-acting DNA elements 5’ to the tran- scription start site. Deletion of the 5’ region of the human insulin gene between nucleotides -279 and -258 caused a 25-fold rise in transcriptional activity whereas further deletion to nucleotide -229 reduced transcription activity 25-fold. In vitro analysis of protein binding in the 5’ regulatory region

Insulin is a polypeptide hormone of vital physiological importance in the control of glucose homeostasis in animals (Steiner and Tager, 1979). It is synthesized exclusively in /? cells of the islets of Langerhans.
Tissue-specific expression of the insulin gene has been shown to be controlled by 5'flanking regulatory regions of DNA (Walker et al., 1983;Bucchini et al., 1986;Selden et al., 1986).
Functional characterization of &-acting regulatory DNA regions has shown that distinct classes of regulatory elements exist, depending on their ability to activate or repress a given gene, and whether their activity is affected by changes in position and orientation (reviewed in Laimins et al., 1983;Maniatis et al., 1987). These elements are mosaics of binding motifs for protein tram-acting factors with differing cell-type specificities and transcriptional properties (Maniatis et al., 1987;Nomiyama et al., 1987;Ondek et al., 1987). In combination, and when bound to their cognate DNA motifs, these truns-acting factors determine the temporal and spatial expression pattern of the gene and its inducibility by various external stimuli, such as second messenger molecules (Montminy et al., 1986;Angel et al., 1987;Imagawa et al., 1987) or steroid hormones and related ligands (Klock et al., 1987;Glass et al., 1987).
Studies on rat insulin I gene have demonstrated the presence of cell-specific enhancer and promoter elements (Edlund et al., 1985). Further mutational studies have defined regions within the enhancer and promoter which contribute to transcriptional activity, in particular two motifs which bind a single factor, IEF-1, which are essential for /3 cell-specific transcriptional activity (Moss et al,, 1988;Karlsson et al., 1987Karlsson et al., , 1989. Protein-binding studies indicate that numerous other cell-specific and ubiquitous factors bind to the rat insulin I 5' region, but their function has so far remained undetermined (Ohlsson and Edlund, 1986;Moss et al., 1988). Although substantial studies have given an insight into the regulation of the rat insulin I (Moss et al., 1988;Karlsson et al., 1987Karlsson et al., , 1989 and rat insulin II genes (Crowe and Tsai, 1988;Whelan et al., 1989), only preliminary studies upon the 5' regulatory region of the human insulin gene have been performed, indicating that sequences up to -258 were sufficient to direct cell-specific expression (Walker et al., 1983).
In this study, we have undertaken an investigation into the molecular mechanism of the tissue-specific regulation of the human insulin gene. To accurately characterize the 5' regulatory regions of the human insulin gene, we constructed nested 5' deletion mutants from nucleotides -339 to -169 and investigated their ability to control expression of a reporter gene, chloramphenicol acetyltransferase (CAT),' in insulin producing cells and non-producing cells. Deletion of the 5' region of the insulin gene from -279 to -258 gives rise to a 25-fold increase in transcriptional activity whereas further deletion to -229 results in a 25-fold decrease. Gel retardation/methylation interference analysis and DNase footprinting shows that there are multiple binding sites for protein factors with varying cell specificities between -278 and -77. Two of these motifs GAGA (-268 to -278) and GC-II (-234 to -239) contribute toward negative and positive regulation of the human insulin gene, respectively. The GAGA motif can down-regulate a heterologous promoter, but only in p cells and may thus represent a novel class of &-acting element. Construction-Plasmids were constructed using standard methods (Maniatis et al., 1982). 5' deletion mutations were generated from a PstI fragment (-339 to +112) of the human insulin gene cloned (Bell et al., 1980) (Gorman et aZ., 1982;Clark et al., 1989). End-points of Ba/31 deletions were mapped by chemical sequencing (Maxam and Gilbert, 1980) or dideoxy sequencing (Sanger et al., 1977) (Graham and Van der Eb, 1973). The CAT assay was performed as described by Gorman (1985). DNA Labeling-The probe for DNase footprinting was labeled at either end using unique BamHI or Hind111 sites in the polylinker of the plasmid pUins302. After linearization, phosphatase treatment, and purification, DNA was end-labeled using [32P]ATP (Amersham Corp., 3000 Ci/mmol) and T4 polynucleotide kinase as described by Maniatis et al., (1982). Following redigestion with a second enzyme, the labeled probe was separated by polyacrylamide gel electrophoresis and eluted as described by Maxam and Gilbert (1980). Oligodeoxynucleotide (oligo) probes were labeled and annealed as previously described . Preparation of Nuclear E&acts-Nuclear extracts were prepared as described by Dignam et al., (1983)  Positions of point mutations within mutant oligos are underlined. incubated at 37 "C for 15 min, 90 "C for 2 min, then phenol/chloroform extracted, ethanol precipitated, and run on a 8% polyacrylamide, 7 M urea sequencing gel alongside a G+A chemical sequencing ladder of the probe (Maxam and Gilbert, 1980). Gel Retardation and Methylation Interference Adysis-Gel retardation and methylation interference analysis of DNA-protein interactions were performed exactly as described elsewhere (Beam and Docherty, 1989).

RESULTS
Deletion Analysis of the Insulin Gene 5' Region-Previous studies have shown that a minimum of 258 bp of 5'-flanking insulin DNA is sufficient to confer transcriptional activity in p cell lines (Walker et al., 1983). To accurately map the minimum length of 5'-flanking DNA necessary to confer cellspecific activity of a reporter gene (CAT), nested deletions between nucleotides -339 and -169 were constructed using Ba131 nuclease and restriction sites within this region (Fig.  1A). These constructs were transfected into HIT M2.2.2 cells and cell extracts assayed for CAT activity (Fig. 1B). Data from CAT assays were normalized against fi-galactosidase activity from a cotransfected control plasmid pRSV@GAL (Edlund et al., 1985).
In agreement with previous reports (Walker et al., 1983), the largest insulin 5' construct tested (pBCins-339) conferred a CAT activity of 4% of that obtained with a Rous sarcoma virus long terminal repeat (pRSVCAT) used as a positive control. However, deletion from -279 to -258 caused a large rise in CAT activity (Fig. 1B). Further deletion to -229 caused a decrease of CAT activity to levels seen with the -339 construct.
Deletion to -201 caused a further fall of CAT activity whereas deletion to the BglII site at -169 caused a small rise in CAT activity. Although the variations of CAT activity caused by deletion of sequences 3' to -229 are relatively small compared with the activity of the ins-258 construct, the changes were highly reproducible and above basal CAT activity from the promoterless vector pBC0. These results suggest that deletion of sequences of insulin 5' DNA removes sites for both positive and negative trans-acting factors.
When transfected into Cos-7 and BHK cells none of these constructs generated amounts of CAT activity significantly .a above levels generated by the promoterless control pBC0 (results not shown).
Multiple Trans-acting Factors Bind to the 5' Region of the Human Insulin Gene, within Regions Identified to Control Transcription-To investigate the contribution of DNA-binding protein factors to the transcriptional activity of the human insulin regulatory region, gel retardation and methylation interference analysis and DNase footprinting using crude nuclear extracts derived from insulin-producing HIT M2.2.2 cells and from heterologous cell lines which do not express the insulin gene (HeLa, BHK, HL60, and Jurkat) were employed. Double-stranded overlapping oligo probes, spanning insulin 5'-flanking regions between nucleotides -289 and -171 (Fig. 2) were synthesized and used as probes in gel shift assays. Probes B, C, D, E, and N all generated multiple or single retarded bands when incubated with HIT M2.2.2 nuclear proteins ( Fig. 3A; see below for oligo(N)-binding data). The complexes formed by protein binding to probes B, C, and D were abolished by performing the reactions in the presence of excess cold unlabeled oligo (B), (C), and (D), respectively, but not with heterologous competitors (Figs. 3B, 4A, and 5A). The weak band seen with probe E did not respond to any competitors including E (results not shown).

The CT-binding
Site-Methylation interference analysis was used to demonstrate that the major complex formed by oligo (B) resulted from binding of a protein factor to the sequence CTAATG (CT-II motif) between nucleotides -215 and -210 (Fig. 3C). This binding motif and characteristic gel retardation pattern suggest that the CT-II-binding factor is identical to that observed in a RINm5F (rat insulinomaderived cell line) nuclear extract, termed insulin upstream factor 1 (IUF-1) . To confirm the similarity of this complex to that observed in RINm5F cells, two mutant variants of oligo (B) (Bml and Bm2, see analysis of the closely spaced doublet generated by a factor, present in HIT M2.2.2 cells, with oligo (C). Residues which interfere with protein binding are marked 0. 0 denotes partial interference. B, gel retardation analysis of the tissue-specific distribution of CRE-binding factors in HIT M2.2.2 cells (lanes 1 -and 2) and heterologous cell lines: HIT T15-G (lanes 3 and 4), BHK (lanes 5 and 6), HeLa (lanes 7 and 8), HL60 (lanes 9 and IO), and Jurkat (lanes II and 12) using labeled probes oligo (C) (lanes 1,3,5,7,9,11) and oligo (Cm) (lanes 2, 4, 6, 8, 10, 12). tion -216, competed efficiently for binding to the CT-II motif in oligo (B) and also formed a complex of identical mobility to that seen with oligo (B) (results not shown). Oligo (Bm2), which contains a single A to C mutation at position -212, was inactive as a competitor for binding to the CT-II motif (Fig. 30, lane 3). These data are qualitatively identical to previous observations using RINm5F nuclear extracts .

Regulation of the Human Insulin Gene
To determine the tissue distribution of IUF-I, gel retardation analysis of nuclear extracts from heterologous, noninsulin-producing cell lines (HL60, BHK, HeLa, Jurkat) was performed with oligo (B). Oligo (B) was not retarded by HL60 or BHK cell extracts (data not shown). The retarded bands formed by HIT.T15-G nuclear extract were of identical mo-  Complexes formed by HeLa and Jurkat nuclear proteins with oligodeoxynucleotide B were of different mobility to those seen with HIT M2.2.2 nuclear extracts. The major complexes formed by HeLa cell nuclear proteins were not abolished by excess unlabeled oligo (B) or (Bm2) (Fig. 30, lanes 7-9) and therefore probably arise by a nonspecific interaction.
In Jurkat cells, competition analysis with oligos (B) and (Bm2) indicated that binding of factors to oligo (B) is unlikely to involve the CT-II motif, since binding is abolished by excess cold oligo (Bm2) (Fig. 30, lanes 10-12). Thus, characteristic binding to the CT-II motif was seen only in HIT M2.2.2 cells, that express the insulin gene, or in HIT-T15-G cells.
The CREB-binding Site-Methylation interference analysis of the closely spaced doublet retarded band formed by oligo (C) (Fig. 4A) revealed binding to a sequence TGACG between nucleotides -179 and -183 (Fig. 4B). This sequence is identical to the half-site binding sequence of the CREBfactor (Fink et al., 1988) which mediates transcriptional activation in response to stimuli which raise CAMP levels. Further 3' to this CRE motif there is divergence from the known CRE motif, TGACGTCA. By using a mutant oligo (C) with an inactive "CRE" (oligo(Cm), see Fig. 2) which fails to bind to or compete for complexes formed by oligo (C) within the CRE-like motif, it was possible to verify that all the cell lines tested in this study contain factors which bind specifically to the CRE-like sequence in the insulin gene 5' region (Fig. 4C). This wide distribution profile might be expected for an important modulatory protein such as CREB.
The GC-binding Site-Methylation interference analysis of the specific complex formed by oligo (D) (Fig. 5A) showed binding of protein to a sequence GCCACC between nucleotides -239 and -234 (GC-II motif) (Fig. 5B). The GC-II motif lies within the boundaries of the region mapped by deletion analysis known to contain a powerful positive regulatory element (see Fig. 1B). Furthermore, the position of the GC-II motif and similarity to the sequence GCCATCTG known to be important for rat insulin I gene transcriptional activity (Moss et al., 1988, Karlsson et al., 1989 suggests that the GC-II motif may act as a binding site for the rat insulin I gene regulatory factor IEF-1 ( (Fig. 2) were used in competition gel shift assays (Fig. 6). Oligo (Dm2) contains a G to T transversion at nucleotide -233. This generates a motif GCCACCTG, similar to the consensus NF-PE binding motif present in the IgH enhancer (Ephrussi et al., 1985) but within the context of the human insulin gene regulatory region. Mutant Dm3 contains a C to T mutation at position -235 as well as the nucleotide change previously described for mutant Dm2. This generates the sequence GCCATCTG, identical to the proposed IEF-l-binding motif present in the rat insulin I gene (Moss et al., 1988, Karlsson et al., 1989 and at a similar position. This motif is also present in the human insulin gene between nucleotides -104 and -111 (GC-I motif) (Fig. 6A). Mutant oligos (Dm2) and (Dm3) were able to compete more efficiently than wild-type oligo (D) for binding to the GC-II motif (Fig. 6B). When oligo (Dm2) and (Dm3) were used as probes they formed complexes of identical mobility to the complex formed by oligo (D). Furthermore, these complexes formed by oligos (Dm2) and (Dm3) were able to be cross- competed for by each other and by oligo (D) (Fig. 6C). These data suggest that the same factor binds to the GC-I motif, the GC-II motif, and the NF-PE-like motif present in oligo (Dm2).
To determine the tissue distribution of the GC-II-binding factor, nuclear extracts from HIT.Tl&G, BHK, HL60, HeLa, and Jurkat were analyzed in a gel shift assay with oligo (D) as a probe. To test whether complexes formed arose as a result of interaction with the GC-II motif, binding to another mutant variant of oligo (D), (Dml, which contains a C to A transversion at nucleotide -237 (Fig. 2)) was studied. This mutation blocked binding to the GC-II motif (Fig. 5C, lanes 1  and 2). Factors that bound to the GC-II motif were detected in nuclear extracts from all cell lines tested, except for BHK cells (Fig. 5C).
The GG-binding Sites-DNase footprint analysis of HIT M2.2.2 nuclear proteins showed protection of the CT-I-  and CT-II-binding sites (Fig. 7,  F2 and F4), the GC-I and GC-II sites (Fig. 7, not indicated), and the negative regulatory element (see below; Fig. 7, Fl). In addition a binding motif between nucleotides -127 and -153, (Fig. 7, F3) not detected previously by gel shift analysis, was protected. Footprint F3 covers a region containing the motif GGAAAT in tandem repeat separated by a small palindromic pentamer motif that seems to be unprotected. Analysis of binding to the GG motifs by factors in other cell extracts showed that the GG-I motif is protected by factors present in heterologous cells, but the GG-II motif is protected only in the insulin-producing HIT M2.2.2 cell line (Fig. 7B). We surmize that each of these tandem repeated motifs, GG-I and GG-II (Fig. 10) comprises a discrete binding site, possibly for the same factor, although the differential footprinting of these motifs by the heterologous cell extracts would militate against this, unless the two sites bound the same factor with widely differing affinities.
A Powerful Negative Regulatory Element in the 5' Regulatory Region of the Human Insulin Gene-Deletion of nucleotides -279 to -258 in the human insulin gene caused a large (25-fold) rise in transcription of the CAT reporter gene (Fig.  1B). This was taken to represent deletion of a binding site for a trans-acting factor capable of suppressing the positive transcription-regulatory activity associated with sequences 3' to position -260. To investigate further the properties of this negative regulatory element (NRE), a 340-bp PuuII fragment derived from pUins301 and containing insulin 5' sequences -3396 to -258 was inserted in both orientations upstream of the TK promoter in the vector pBCTKp-1. This fragment of the human insulin gene caused a 50% decrease in TK promoter activity in HIT M2.2.2 in the sense orientation; in the antisense orientation only a very slight effect upon TK promoter activity was observed (Fig. 8). In contrast, a 200-bp PuuII/HindII fragment of the human insulin gene (nucleotides -258 to -60) (Fig. 2) caused a marked enhancement of TK promoter activity (Fig. 8). These effects were not observed in Cos-7 cells (data not shown). The NRE is situated between nucleotides -279 and -258, near the 3' end of the P&I fragment used in the TK promoter experiments, thus, reversal of orientation of this fragment moves the NRE about 320 bp further 5' as well as reversing orientation. Therefore, this A, a PstI/HindII fragment of the insulin gene subcloned into pUC18 (pUins302) was 5' end-labeled at either end using unique sites within the polylinker and used as a probe in the DNase I footprint assay. Resolution of sequences distal to the labeling site was achieved by running gels for a longer period and using multiple loadings of the same sample. B, tissue-specific distribution of factors that bind to the GG motifs in the insulin gene 5' region, using DNase footprint analysis. A DNA probe labeled at -59 was incubated with 50 pg of the following ex-  (Laimins et al., 1986;Saffer and Thurston, 1989;Burt et al., 1989), since it is sensitive to changes in position and orientation.
Also, it can directly down-regulate a heterologous promoter without requirement of a functional enhancer, distinguishing it functionally from the porcine MHC class 1 gene negative regulatory region (Ehrlich et al., 1988).
The putative insulin NRE contains a dyad symmetry element GGAGA flanked with A. T triplets either side in tandem repeat (Figs. 1 and 10) between nucleotides -274 and -285. Dyad symmetry elements within regulatory DNA regions have been shown to be candidate binding sites for trans-acting factors (Montminy et al., 1986;Angel et al., 1987;Klock et al., 1987). To investigate protein binding to this region, we performed gel shift analysis of proteins binding to an oligo spanning the putative insulin NRE, oligo (N) (-260 to -289) (Fig. 2). Oligo (N) formed multiple DNA-protein complexes with factors present in HIT M2.2.2 cells. These retarded complexes were competed for specifically by oligo (N) (Fig.  9A). Methylation interference analysis of complexes B2 to B4 shows that these complexes seem to arise from the binding of nuclear factors, present in HIT M2.2.2 extracts, to three distinct overlapping binding sites, as determined by the differential interference patterns of the G and A residues within the binding sites (Fig. 9B). The overlapping binding motifs are detailed in Fig. 10. DNase footprinting of HIT M2.2.2 cell nuclear extract also revealed a protected region over the protein-binding sites within oligo (N) (Fig. 7A, footprint Fl).
Gel retardation analysis of proteins present in nuclear extracts from heterologous cell lines with oligo (N) revealed that most of the bands present when a HIT M2.2.2 extract are used are present in other cell lines but only as subsets of the pattern seen with HIT M2.2.2 cell extracts. Only band B4 is represented exclusively in HIT M2.2.2 cells (Fig. SC) p cell-specific, therefore band B4 may be implicated in the tissue-specific function of the NRE.

DISCUSSION
In this investigation we have shown that 5'-flanking regions of the human insulin gene modulate activity of a fused reporter gene (CAT) in a /3 cell-specific manner. Deletions between nucleotides -219 and -229 alternatively negatively and positively modulate CAT activity in a fi cell-specific manner. Analysis of protein binding for the 5' region of the human insulin gene and within areas delineated as being important transcriptional control elements shows the binding of multiple cell-specific and ubiquitous nuclear protein factors ( Fig. 10 and Table I). In particular, binding sites for multiple protein factors have been located within the powerful positive and negative regulatory regions of the human insulin gene 5' DNA.
Protein-binding Sites within the Regulatory Region of the Human Insulin Gene-This study has shown that the human insulin gene regulatory region contains many binding sites for both highly cell-specific and ubiquitous protein factors, some of which bind to multiple sites and some of which bind to overlapping sites. Previous studies on the rat insulin gene protein binding to the enhancer have been reported (Ohlsson and Edlund, 1986;Ohlsson et al., 1988;Moss et al., 1988).
Because of the 70% similarity between regulatory regions of the rat and human insulin genes, it might be expected that they contain similar binding motifs. We have found this to be the case in some instances, but we find some of our data contradictory to that reported by previous workers. We have shown here and elsewhere  binding of an islet cell-specific factor to the sequence motif CTCTAATG (CT-II motif) between nucleotide -217 and -210. DNase footprint analysis also shows protection of a related motif (CT-I) between nucleotide -77 and -84. These observations confirm those made in an earlier study on the RINm5F islet /3 cell line . The CT-II motif is similar in sequence and relative position to the Ez motif in the rat insulin gene defined by DNase footprinting (Ohlsson and Edlund 1986;Ohlsson et al., 1988).
We have detected, using gel shift/methylation interference analysis, binding of a nuclear factor to a motif (GC-II) with 75% similarity, and in a comparable location, to the IEF-1 binding site (Moss et al., 1988, Karlsson et al., 1989. It has been proposed that the IEF-1 factor, which binds at two locations in the rat insulin gene, is the dominant @ cellspecific positive truns-activating protein for insulin gene expression (Karlsson et al., 1989, Karlsson et al., 1987. Our studies indicate that it is highly probable that IEF-1 binds to the proximal and distal GC-I and GC-II motifs which we have defined in the human insulin gene. The pronounced fall in transcriptional activity seen when the GC-II motif was deleted would also tend to support this line of reasoning. We have shown that factors which bind to the GC-II motif are not confined to /3 cell lines but are also present in other cell lines. Whether these factors are related to IEF-1 or to any of the NF-PE factors (Moss et al., 1988) remains to be determined.
Two recent studies on the rat insulin II gene appear to contradict each other as to whether the IEF-l-binding site is important in the control of rat insulin II gene expression (Crowe and Tsai, 1989;Whelan et al., 1989).
We have noted that the GC and CT motifs are paired to each other with a 15-bp intervening region (Fig. 10) at two positions in the insulin 5' region. Karlsson et al. (1987) have shown that the corresponding rat insulin I CT-II motif, although devoid of intrinsic transcriptional activity, serves to potentiate activity of the GC motif, perhaps by facilitating long range interactions.
This duplication of GC/CT motifs seen in the human insulin gene leads us to suggest that close juxtaposition of CT and GC motifs constitutes a /3 cell-specific protoenhancer, similar to the interaction between A and B domains of the SV40 enhancer (Zenke et al., 1986;Ondek et al., 1987).
We have identified a protein-binding site between nucleotide -179 and -183 which has 75% identity with the binding site for the CREB factor (Montminy et al., 1986;Fink et al., 1988). CREB is a transcription factor whose activity is modulated through the cyclic AMP/protein kinase A pathway (Montminy and Bilezikjian, 1987;Yamamoto et al., 1987).
Thus, cyclic nucleotide regulatory elements can directly control expression of a CAMP-responsive gene in relation to stimuli which alter levels of cellular CAMP. Insulin gene transcription is modulated by changes in cellular CAMP, and it has been proposed that glucose elevates insulin gene expression in islets via a CAMP-mediated pathway (Nielson et al., 1985). We are currently investigating whether the "CRE" element in the insulin gene 5' region is functional in the modulation of glucose-mediated changes in insulin gene expression.
DNase protection studies have revealed a protected region of the insulin gene 5' region between nucleotides -127 and -153 composed of two GGAAAT (GG) motifs in tandem repeat with a pentanucleotide dyad symmetry between them. It is worthy of note that only differences in the protection pattern of the GG-II motif distinguish between the proteinbinding profile of HIT M2.2.2 cells and HIT.T15-G cells. Whether this is functionally significant with respect to insulin gene expression remains to be determined.
Although a GGlike motif exists in the rat insulin I gene, footprinting of the rat insulin I gene enhancer showed no reproducible footprint over the GG motif (Ohlsson and Edlund, 1986), and selective I T -278 G &Imutagenesis of this region in the rat insulin I gene did not decrease its transcriptional activity (Karlsson et al., 1987). When the corresponding GG-like motif of the rat II insulin gene 5' region was mutated, this led to a loss of transcriptional activity (Crowe et al., 1989). Although the human GG motifs have not been studied functionally, it is clear from comparative studies on rodent insulin genes that there is some redundancy of protein-binding sites built into the insulin gene 5' regions.
Negative Regulation of the Human Insulin Gene-Recently, it has become increasingly obvious that control of eukaryotic gene transcription involves negative as well as positive regulation (Goodbourn et al., 1986;Baniahmad et al., 1987;Gaub et al., 1987;Imler et al., 1987;Reue et al., 1988;Burt et al., 1989;Ehrlich et al., 1989). Previous studies on the rat I, rat II, and human insulin genes have demonstrated that they are negatively regulated in non-P cell lines (Nir et al., 1986;Laimins et al., 1986;Takeda et al., 1989;Whelan et al., 1989). In the human insulin gene the 5' polymorphic region may be involved in negative regulation (Takeda et al., 1989) implicated. Thus, it appears that a unified negative regulatory mechanism for insulin genes does not appear to exist. In this study we have shown that negative regulation contributes to activity of the human insulin gene regulatory region in a /3 cell-specific manner. The NRE we have described differs markedly from others in its strength and in its apparent p cell specificity (although other NREs associated with insulin genes have not been rigorously tested in insulin secreting cell lines). Since the GAGA box is inactive in non-b cell lines under the conditions tested it appears not to play a role in regulating the specificity of insulin gene expression, as has been suggested for other negative regulators. There is no evidence for a similar p cell-specific NRE in the rat insulin I gene (Walker et al., 1983, Karlsson et al., 1987. We have noted that a similar deletion study on the 5' region of the human insulin gene failed to show anything more than a 20% rise in transcriptional activity when sequences between -339 and -258 were deleted (Walker et al., 1983). One possible explanation for this discrepancy may have been the use of different vectors for the CAT assays. However, we have faithfully repeated our observations using insulin 5' deletions shown). Other possible explanations of the p cell type-specific negative regulatory effect observed in this study are that: (a) this effect results from an artifact due to the use of a species heterologous system (i.e. human insulin gene and a hamster ,8 cell line). It should, however, be pointed out that most of the studies on the rat insulin I and II genes were also performed on the same cell line. This hypothesis cannot be adequately investigated until human p cell lines become available. (b) The activity of the NRE in down-regulating the positive element between -260 and -229 is compensated for by the presence of another enhancer element located elsewhere in the insulin gene. High activity of human insulin genes bearing 2-5 kilobases of 5'-and 3'-flanking region, in transgenic mice (Selden et al., 1986, Bucchini et al., 1986 suggest that this may be a worthwhile area of investigation. (c) The NRE may play a physiological role in down-regulating the insulin gene in response to absence of a critical factor or constituent from the culture medium. Different culture conditions may also account for the discrepancy between our data and those of Walker et al. (1983).
Recent studies have shown that many regulatory regions contain alternating positive and negative regulatory elements. Based upon these studies and our own, we propose two models to describe how these non-enhancer-like negative elements may function. (a) General interaction: negative and positive elements exert their effects separately. Net residual enhancer activity results after a generalized cancellation of negative and positive effects. (b) Specific interaction: negative elements specifically down-regulate designated positive elements to which they are correctly juxtaposed. Residual enhancer activity is contributed by remaining "uncancelled" regulatory elements.
We surmize that the latter model more accurately explains our observations because of the position and orientation sensitivity of the insulin negative element. Further experimentation is needed to elucidate the levels of interaction between individual regulatory elements by making chimeric constructs bearing different combinations of negative and positive elements.

trans-acting factors.
Positive and negative regulation of the human insulin gene by multiple