Differential Effects of Heterogeneous Nuclear Ribonucleoprotein K on Sp1- and Sp3-mediated Transcriptional Activation of a Neuronal Nicotinic Acetylcholine Receptor Promoter*

The neuronal nicotinic acetylcholine receptor gene family consists of 11 members, α2–α9 and β2–β4. Three of the genes, those encoding the α3, α5, and β4 subunits, are clustered tightly within the genome. These three subunits constitute the predominant acetylcholine receptor subtype expressed in the peripheral nervous system. The genomic proximity of the three genes suggests a regulatory mechanism ensuring their coordinate expression. However, it is likely that gene-specific regulatory mechanisms are also functioning because the expression patterns of the three genes, although similar, are not identical. Previously we identified regulatory elements within the β4 promoter region and demonstrated that these elements interact specifically with nuclear proteins. One of these elements, E1, interacts with the regulatory factor Purα as well as three other unidentified DNA-binding proteins with molecular masses of 31, 65, and 114 kDa. Another element, E2, interacts with Sp1 and Sp3. Because E1 and E2 are immediately adjacent to one another, we postulated that the proteins that bind to the elements interact to regulate β4 gene expression. Here we report the identification of the 65-kDa E1-binding protein as heterogeneous nuclear ribonucleoprotein K and demonstrate that it affects the transactivation of β4 promoter activity by Sp1 and Sp3 differentially.

Key molecular components of synapses are ligand-gated ion channels that are intimately involved in generating the electrical signals that underlie information processing within the nervous system. The most well characterized ion channels are the members of the nicotinic acetylcholine (nACh) 1 receptor family. Sixteen genes encoding members of this family have been identified, with 11 of them, ␣2-␣9 and ␤2-␤4, being expressed within neuronal populations (1)(2)(3)(4)(5). Reconstitution studies have shown that the ␣2-␣6 subunits in combination with either the ␤2 or ␤4 subunit can each lead to the formation of functional nACh receptors. Each of these receptor subtypes has distinct electrophysiological and pharmacological properties (4,6). In contrast, the ␣7-␣9 subunits appear to be able to form homomeric receptors (7)(8)(9). It is likely that the functional diversity exhibited by the neuronal nACh receptor family results from the differential expression of these subunits which leads to incorporation of different subunits into mature receptors. Although the consequences of this diversity are beginning to be appreciated, an understanding of the molecular basis governing expression of neuronal nACh receptor subunits remains elusive. However, recent advances indicate that regulation of the receptor subunit genes at the level of transcription plays a critical role (see below).
In situ hybridization studies have demonstrated that each of the nACh receptor subunit genes exhibits distinct temporally and spatially restricted patterns of expression in the peripheral and central nervous systems, suggesting that they are most likely regulated independently (although in some cases there is overlap in expression patterns which may indicate that such genes share certain regulatory features, as discussed below; Ref. 2, 3, 10 -15). As mentioned above, it is clear that transcriptional regulation, both positive and negative, plays a key role in the establishment of the differential expression patterns of the subunit genes (16 -36). We would like to understand the molecular details of this regulated expression. We have focused upon characterizing the transcriptional mechanisms involved in the expression of a cluster of receptor subunit genes, those encoding the ␣3, ␣5, and ␤4 subunits. This cluster of genes spans approximately 60 kilobase pairs of the rat genome (see Fig. 1 and Ref. 37). Because the ␣3, ␣5, and ␤4 subunits make up the predominant nACh receptor subtype expressed in the peripheral nervous system (38,39), the clustering of their genes raises interesting questions regarding the regulatory basis of their coexpression. It is certainly plausible that these genes are expressed coordinately via a set of common regulatory mechanisms. On the other hand, neither the temporal nor the spatial patterns of expression of the ␣3, ␣5, and ␤4 subunit genes are completely identical. Although the developmental and functional implications of these temporal and spatial differences in gene expression are unclear, they underscore the conclusion that each gene of the cluster has unique aspects to its regulation.
Previously we identified several regulatory elements within the promoter region of the ␤4 gene and demonstrated that these elements interact specifically with nuclear proteins present in extracts prepared from brain tissue and an established neuronal cell line, SN17 (22,30). One of these elements, E1, interacts with the transcriptional regulatory factor Pur␣ (34) as well as three other unidentified DNA-binding proteins that we refer to as neuronal ACh receptor promoter-binding proteins (NARP). These proteins have molecular masses of 31, 65, and 114 kDa (34). Another element, E2, interacts with the transcription factors Sp1 and Sp3 (30,31). An intact E2 is required for transcriptional activation of the ␤4 promoter by Sp1 and Sp3 (31). Because E1 and E2 are immediately adjacent to one another and both are required for wild type ␤4 promoter activity, we postulated that the proteins that bind to the elements interact to regulate ␤4 gene expression (31). In this report, we describe the identification of the E1-binding protein NARP65 (34) as heterogeneous nuclear ribonucleoprotein K (hnRNP K) and demonstrate that it affects the transactivation of ␤4 promoter activity by Sp1 and Sp3 differentially.

EXPERIMENTAL PROCEDURES
Protein Purification and Peptide Sequencing-The E1-binding protein, NARP65, was purified from crude nuclear extracts prepared from bovine brains as described previously (34) with some modifications. Briefly, after partial purification by P11 phosphocellulose chromatography, fractions containing E1 binding activity were pooled, concentrated, and dialyzed for further purification by two rounds of DNA sequence-specific affinity chromatography. Two 20-base complementary oligonucleotides containing the E1 sequence (5Ј-GACCCTC-CCCTCCCCTGTAA-3Ј; see Fig. 1) were used to make the DNA affinity matrix. Annealed, concatenated oligonucleotides were coupled to cyanogen bromide-activated Sepharose 4B (Amersham Pharmacia Biotech) essentially as described by Kadonaga and Tjian (40). After the DNA sequence-specific affinity chromatography, fractions containing E1 binding activity were pooled, precipitated with acetone, and analyzed on 10% SDS-polyacrylamide gels. After staining with Brilliant Blue R-250 (Fisher Biotech) and destaining, NARP65 was excised, washed with 50% methanol, and subjected to amino acid microsequencing (Harvard Microsequencing Facility) as described previously (41).
Substrate Preparation for Electrophoretic Mobility Shift Assay (EMSA)-The E1/E2 oligonucleotides ( Fig. 1) used as probes and competitors in the EMSA were synthesized using an Oligo 1000 DNA Synthesizer (Beckman). Single-stranded oligonucleotides corresponding to either the coding or noncoding strand of E1/E2 were labeled radioactively with [␥-32 P]ATP (NEN Life Science Products) using T4 polynucleotide kinase (Promega). When double-stranded probes were used, equal amounts of E1/E2 complementary oligonucleotides were annealed and then labeled radioactively. The specific activities of these probes were typically between 5,000 and 15,000 cpm/fmol.
Protein Expression and Preparation of Nuclear Extracts from Transfected Drosophila Cells-A glutathione S-transferase (GST)-hnRNP K fusion construct was generously provided by Dr. David Levens (National Cancer Institute). Expression and purification of GST and GST-hnRNP K fusion proteins were carried out as described previously (42). Nuclear extracts from transfected Drosophila cells were prepared by the method of Dignam et al. (43) as described previously (44) except that nuclear extracts were dialyzed against binding buffer of composition 50 mM Tris-HCl (pH 7.5), 50 mM NaCl, 5 mM MgCl 2 , 1 mM dithiothreitol, 1 mM EDTA, 5% glycerol, 100 M ZnCl 2 .
EMSA-When using recombinant proteins, EMSAs were performed as described previously (22) with radiolabeled double-or singlestranded oligonucleotides incubated with the indicated amounts of GST or GST-hnRNP K in the presence of 2 g of nonspecific competitor poly(dI-dC). For competition experiments, unlabeled double-or singlestranded oligonucleotides were preincubated with the recombinant proteins for 15 min before the addition of labeled oligonucleotides. After a 1-h incubation at 4°C, the reaction mixtures were electrophoresed through 6% native polyacrylamide gels and visualized by autoradiography. When using nuclear extracts as protein sources, EMSAs were performed using a 32 P-labeled double-stranded E1/E2 oligonucleotide ( Fig. 1) and 3.5 g of nuclear extracts prepared from transfected SL2 cells. Reaction mixtures containing nuclear extracts, binding buffer, and 2 g of poly(dI-dC) were preincubated for 5 min at room temperature before the addition of 5 fmol of the end-labeled probe. After addition of the probe, binding reactions were incubated further for 15 min at room temperature. Reaction mixtures were then electrophoresed through 6% native polyacrylamide gels. Radioactivity was detected by autoradiography of the dried gels.
Expression Constructs-The Sp1 expression construct pActSp1, containing the Sp1 coding sequence inserted downstream of the Drosophila melanogaster actin 5C promoter, was generously provided by Dr. Ed Seto (University of South Florida; Ref. 45). The Sp3 expression construct pPacUSp3, containing the Sp3 coding sequence inserted downstream of the actin 5C promoter, was kindly provided by Dr. Guntram Suske (Philipps-Universitä t Marburg, Germany; Ref. 46). The pAct-hnRNP K expression plasmid was constructed by subcloning the mouse hnRNP K coding sequence from pCRII-hnRNP K (the kind gift of Dr. Karol Bomsztyk, University of Washington) into the pAct vector. The EcoRI fragment containing the mouse hnRNP K coding sequence was isolated from pCRII-hnRNP K, treated with the Klenow fragment of DNA polymerase I and ligated subsequently to EcoRV-digested pAct to generate the construct pAct-hnRNP K.
Cell Culture and Transfections-SN17 cells (47) were cultured as described previously (22). D. melanogaster Schneider SL2 cells were obtained from the American Type Culture Collection and were maintained at room temperature in modified Schneider's Drosophila medium (Life Technologies, Inc.) supplemented with 10% fetal bovine serum and antibiotics. Cells were seeded at 2 ϫ 10 6 /35-mm culture well immediately before transfection. DNAs were introduced into the cells by liposome-mediated transfection using 9 l of CellFECTIN/sample (Life Technologies, Inc.) and 0.75 g of target DNA (pX1B4FHwt, see The positions of two transcriptional regulatory elements, E1 and E2, within the ␤4 5Ј-flanking region; the DNA sequence encompassing the elements is shown also. The sequence of the upper strand of the E1/E2 oligonucleotide used in EMSA is indicated as E1/E2 oligo. pX1B4FHwt denotes a luciferase expression vector containing the FokI to HindIII fragment as indicated. kbp, kilobase pairs. and Sp3 coding sequences, respectively, as negative controls for luciferase expression. All transfections included 0.23 g of a ␤-galactosidase expression vector, RSV-␤Gal, in which expression of the bacterial lacZ gene is driven by a Rous sarcoma viral promoter. After a 5-h incubation with the DNA-CellFECTIN complexes, the cells were overlaid with 0.5 ml of medium containing 30% fetal bovine serum and incubated for 24 h. The next day, the cells were overlaid with 2 ml of complete growth medium. After another 24 h, the cells were harvested and assayed for luciferase activity using a commercially available kit (Promega Corp.). Luciferase values were normalized to ␤-galactosidase activity, which was measured using a commercially available kit (Galacto-Light; Tropix, Inc.). For nuclear extract preparation, SL2 cells were transfected essentially as described above, except that transfections were carried out in 100-mm culture dishes, and only the effector DNAs were introduced into the cells (pActSp1, pPacUSp3, or pAct-hnRNP K).
Western Blotting-Western blotting was performed essentially as described before (31) with minor modifications. 1 l of an SN17 cell extract and 10 g of nuclear extracts isolated from untransfected or transfected SL2 cells were electrophoresed through SDS-polyacrylamide gels (10%). Anti-hnRNP K antibody 3C2 (the kind gift of Dr. Gideon Dreyfuss, University of Pennsylvania School of Medicine; Ref. 48) was used at a 1:1,000 dilution in Blotto. Experiments utilizing anti-Sp1 and anti-Sp3 antibodies were done as described previously (31).

RESULTS
NARP65 Is the Bovine Homolog of hnRNP K-Our earlier studies led to the identification of two regulatory elements, E1 and E2, located between nucleotides Ϫ54 and Ϫ81 relative to the ␤4 transcriptional initiation site ( Fig. 1; Refs. 21, 22, 30). As mentioned above, E2 is a binding site for members of the Sp family of transcriptional regulatory factors (30,31). Biochemical purification from bovine brain tissue of nuclear proteins that bind to E1 led to the identification, in the most highly purified fraction, of four polypeptides that interact specifically with E1 (34). One of these proteins, NARP43, was shown by amino acid sequence analysis to be the transcriptional regulator, Pur␣ (34). We have now obtained amino acid sequences of two tryptic peptides of another E1-binding protein, NARP65. The sequences of these peptides are identical to two regions of the hnRNP K coding sequence ( Fig. 2; Ref. 48), thus establishing NARP65 as the bovine homolog of hnRNP K.
hnRNP K Interacts Specifically with the ␤4 E1/E2 Region-Recently hnRNP K has been shown to be a transcriptional activator, a function that is dependent upon its binding to a CT element (42,49). hnRNP K binds both single-and doublestranded nucleic acids, having a higher affinity for the former (for review, see Ref. 50). As shown in Fig. 1, E1 is characterized by three repeats of 5Ј-CCCT-3Ј. To determine whether hnRNP K interacts with the E1/E2 region, EMSAs were carried out using a GST-hnRNP K fusion protein and both single-and double-stranded E1/E2 probes. As shown in Fig. 3, hnRNP K interacts specifically with both the double-stranded E1/E2 probe and the upper strand (coding) E1/E2 probe, but not at all with the lower strand (noncoding) E1/E2 probe. As seen most clearly with the coding strand E1/E2 probe, two specific protein-DNA complexes are formed, with the most prominent complex migrating the fastest. The presence of two complexes is most likely a consequence of oligomerization of hnRNP K, as has been reported previously (62). Competition experiments demonstrated that unlabeled double-stranded E1/E2 competed very little for binding to the single-stranded coding E1/E2 probe, whereas unlabeled single-stranded coding E1/E2 virtually eliminated hnRNP K binding to the double-stranded probe (Fig. 3). Unlabeled single-stranded noncoding E1/E2 did not compete for binding to either the single-stranded coding or double-stranded E1/E2 probes, consistent with the observation that the single-stranded noncoding E1/E2 probe did not form any specific complexes with hnRNP K (Fig. 3). These observations indicate that hnRNP K interacts specifically with the E1/E2 region and that the protein appears to have a higher affinity for the single-stranded CT-rich coding E1/E2 probe, as has been reported for hnRNP K binding sites in other genes (50).
hnRNP K Is Expressed Highly in SN17 Cells but Not in Drosophila Cells-To test the potential functional significance of hnRNP K on ␤4 promoter activity, it was necessary to identify a suitable cell line in which to carry out transfection studies. Because high endogenous levels of hnRNP K may complicate interpretation of transfection results, it was desirable to identify a cell line that either does not express hnRNP K or expresses it at very low levels. Two cell lines that we have used extensively in the past to study ␤4 gene expression are the neuronal cell line SN17 and the Drosophila Schneider SL2 cell line (21,22,30,31). To determine the expression levels of hnRNP K in these cell lines, Western blot analysis was carried out. As shown in Fig. 4, SN17 cells express relatively high levels of hnRNP K, but SL2 cells do not express any detectable hnRNP K protein. However, when transfected with an expression construct producing mouse hnRNP K under the control of the Drosophila actin 5C promoter (see "Experimental Procedures"), SL2 cells expressed readily detectable levels of exogenous hnRNP K protein (Fig. 4, lane 3). Therefore, we chose SL2 cells to perform functional analysis of hnRNP K effects on ␤4 promoter activity.
hnRNP K Regulates Sp1 and Sp3 Transactivation of the ␤4 Promoter Differentially-To investigate whether hnRNP K has any effect on the transcriptional activity of the ␤4 promoter, SL2 cells were transfected with a wild type ␤4 promoter/luciferase expression construct (pX1B4FHwt; see Fig. 1) and with increasing amounts of an expression construct for hnRNP K (pActMK) or of the parental vector (pAct) as a control. Surprisingly, given that hnRNP K has been shown to function as a transcriptional activator by virtue of its interaction with a 5Ј-CCCT-3Ј recognition site similar to that of E1 (50), hnRNP K had no effect on ␤4 promoter activity in this assay (Fig. 5).
Our earlier work demonstrated that Sp1 and Sp3 interact with the E2 element, which is immediately adjacent to E1 (Fig.  1) in the ␤4 promoter (see Fig. 1; Refs. 30 and 31). These interactions result in significant activation of the ␤4 promoter in SL2 cells (Refs. 30 and 31; Fig. 5). We hypothesized previously that the proteins binding to E1 and E2 may functionally interact to regulate ␤4 gene expression (30,31). To test whether hnRNP K can affect transactivation of the

FIG. 5. hnRNP K regulates Sp1-and
Sp3-mediated transactivation of the ␤4 promoter differentially. SL2 cells were transfected with pX1B4FHwt (see Fig. 1) alone, with 5 fmol of an Sp1 or Sp3 expression construct separately and together with 5, 10, and 50 fmol of an hnRNP K expression construct (MK), with 5 fmol of an Sp1 or Sp3 expression construct and increasing amounts of MK, with the indicated combinations of all three of the expression constructs, or with the Sp1 expression vector devoid of Sp1 coding sequences (pAct, 50 or 55 fmol, as indicated). Luciferase values were normalized to ␤-galactosidase expression as driven by a Rous sarcoma viral promoter. Fold induction was calculated relative to the normalized luciferase activity obtained by transfecting pX1B4FHwt alone. over background, Sp3 activated the ␤4 promoter 220-fold, and cotransfection of Sp1 and Sp3 resulted in a 440-fold activation of the ␤4 promoter (Fig. 5). Cotransfection of the expression constructs for the Sp factors with the highest amounts of the pAct vector did not have an effect on the transcriptional activities of either Sp1 or Sp3 (Fig. 5). However, cotransfection of hnRNP K with Sp1 resulted in dramatic inhibition of Sp1induced ␤4 promoter activity to background levels even when only 5 fmol of hnRNP K was used (Fig. 5). Interestingly, in the presence of the same amount of hnRNP K (5 fmol) which inhibited Sp1 effects, Sp3 transactivation of the ␤4 promoter/ luciferase construct was reduced only approximately 30% (Fig.  5). Even at the highest amount of hnRNP K (50 fmol) used, Sp3 transactivation of the ␤4 promoter was inhibited only approximately 65% (Fig. 5). When Sp1 and Sp3 were both cotransfected with the ␤4 reporter construct and increasing amounts of hnRNP K, a dose-dependent decrease in transcriptional activity similar to that seen with Sp3 alone was observed (Fig. 5). Thus, despite their similarities, it seems that Sp1 and Sp3 are differentially sensitive to inhibition by hnRNP K.
To test directly whether hnRNP K protein is capable of inhibiting specific DNA binding by Sp1 and Sp3, EMSA experiments were carried out. Nuclear extracts prepared from SL2 cells transfected with the desired combinations of expression constructs were used as protein sources, and E1/E2 oligonucleotide was used as the probe. As shown in Fig. 6, Sp1, Sp3, and hnRNP K exogenously produced in SL2 cells each forms a specific protein-DNA complex of distinct mobility with the E1/E2 elements of the ␤4 promoter. Cotransfection of Sp1 and Sp3 into SL2 cells did not affect the DNA binding activity of either protein (Fig. 6). However, although the ability of hnRNP K to bind to DNA was unaffected by the Sp proteins, the DNA binding activity of Sp1 was virtually abolished, whereas that of Sp3 was only partially reduced when they were cotransfected with hnRNP K (Fig. 6). The presence in the nuclear extracts of the three factors together resulted in the formation of specific protein-DNA complexes similar to those observed with the hnRNP K-Sp3 combination (Fig. 6). These data are consistent with the results of the transactivation experiments presented above. Together they suggest that hnRNP K can repress the transactivation activities of Sp1 and Sp3 by interfering with the DNA binding activities of these proteins. Interestingly, binding to DNA by Sp1 appears to be much more sensitive to inhibition by hnRNP K than that of Sp3.
One possible explanation for the differential affects of hnRNP K is that the Sp1 and Sp3 proteins are differentially sensitive to hnRNP K such that Sp1 protein is degraded more rapidly, through some unknown mechanism, in the presence of hnRNP K. As an initial attempt to address this question, Western blot analysis was carried out using extracts from transfected SL2 cells (Fig. 7). hnRNP K protein expression was unaffected by cotransfection with expression constructs for either Sp1 or Sp3. Sp3 was unaffected by cotransfection with Sp1 and slightly down-regulated by hnRNP K. Sp1 protein levels were somewhat reduced when cotransfected with either Sp3 or hnRNP K. Thus, it appears that Sp1 expression is sensitive to cotransfection with either of the other two proteins, yet in the presence of Sp3, its ability to transactivate the ␤4 promoter is, at the very least, unaffected and may even be enhanced, whereas in the presence of hnRNP K, Sp1 transactivation is eliminated (Fig. 5). DISCUSSION We have begun a molecular dissection of the regulatory processes governing the transcription of neuronal nACh receptor subunit genes with a particular emphasis on the rat ␤4 subunit gene. Four transcriptional regulatory factors have now been identified which interact with the promoter region of the ␤4 gene. Sp1 and Sp3 interact in a functionally relevant manner with E2 (30, 31), whereas Pur␣ and hnRNP K bind to E1 (Ref. 34 and this report). The functional significance of the Pur␣-E1 interaction is currently under investigation. hnRNP K was identified initially as a component of the multiprotein heterogeneous nuclear ribonucleoprotein particle (48), a structure thought to be involved in pre-mRNA processing (51). Although hnRNP K can bind RNA through conserved K homology (KH) domains, its precise role in pre-mRNA processing remains obscure. In addition to binding RNA, hnRNP K has been shown to bind both double-and single-stranded DNA, having a higher affinity for the latter (50). Moreover, in vitro, hnRNP K can stimulate transcription by RNA polymerase II, and in vivo, it can both activate and repress transcription (52). The functional studies on hnRNP K reported here yielded some unexpected results. Surprisingly, hnRNP K by itself did not affect ␤4 pro-FIG. 6. hnRNP K inhibits Sp1, but not Sp3, binding to DNA. EMSAs were performed using 3.5 g of the indicated nuclear extracts, prepared from SL2 cells transfected with various combinations of the "effector" constructs (pActSp1, pPacUSp3, pAct-hnRNP K), and endlabeled, double-stranded E1/E2 oligonucleotide used as a probe. Nuclear extracts were prepared from mock-transfected cells (mock); Sp1transfected cells (Sp1), Sp3-transfected cells (Sp3); hnRNP K-transfected cells (hnRNP K); cells cotransfected with both Sp1 and Sp3 (Sp1 ϩ Sp3); cells cotransfected with both Sp1 and hnRNP K (Sp1 ϩ hnRNP K); cells cotransfected with both Sp3 and hnRNP K (Sp3 ϩ hnRNP K); and cells cotransfected with Sp1, Sp3, and hnRNP K (Sp1 ϩ Sp3 ϩ hnRNP K). Probe indicates a sample in which no nuclear extract was added to the binding reaction. Lettered arrows point to the specific protein-DNA complexes formed; asterisks indicate nonspecific complexes. moter activity in the context of a transactivation assay using Drosophila cells. However, even more surprisingly were the differential effects of hnRNP K on the transactivation abilities of Sp1 and Sp3 on ␤4 promoter activity. In the absence of hnRNP K, both Sp1 and Sp3 transactivate the ␤4 promoter, with Sp3 being the more potent activator. When hnRNP K was included in the transfection, the Sp1-mediated transactivation was virtually eliminated, even at the lowest concentration of hnRNP K used. In contrast, Sp3 transactivation of the ␤4 promoter was repressed only modestly (approximately 30%) by the same low amount of hnRNP K. Even at the highest concentration of hnRNP K used, Sp3 still transactivated the ␤4 promoter 80-fold. Given that all three genes were transcribed from the Drosophila actin 5C promoter, it is unlikely that the differential effects of hnRNP K were a consequence of differential expression of the Sp1 and Sp3 genes. Western blot analysis was used to determine whether the levels of the Sp1, Sp3, and hnRNP K proteins were significantly altered by cotransfection of the various expression constructs. Although Sp1 protein levels were reduced when cotransfected with either Sp3 or hnRNP K expression constructs, it seems unlikely that this is the reason for the differential affects of hnRNP K because cotransfection of Sp3 with Sp1 did not affect the ability of Sp1 to transactivate the ␤4 promoter and, in fact, led to a synergistic activation.
Although the precise mechanism by which hnRNP K exerts its effects on the Sp factors is unknown, it clearly interferes with Sp1 binding to the double-stranded E1/E2 oligonucleotide while having a small effect on Sp3 binding, consistent with the transfection data. Whether these observations reflect different DNA binding affinities of Sp1 and Sp3 in the context of the ␤4 promoter remains to be determined. However, if this were true, a very simple model would suggest that hnRNP K either prevents Sp1 from binding to E2 by physically blocking access to the DNA or displaces bound Sp1 from the DNA. The fact that hnRNP K does not completely block Sp3 function on the ␤4 promoter suggests that Sp3 has a higher binding affinity for E2 than Sp1 such that Sp3 is not displaced from the DNA by hnRNP K. This model obviously highlights differences in the biochemical properties of Sp1 and Sp3 which in turn, most likely underlie the functional distinctions between the two proteins. It is becoming increasingly clear that, despite the significant structural similarities between Sp1 and Sp3, they can differ dramatically in terms of transcriptional regulatory properties. For example, whereas Sp1 has been shown to be a transcriptional activator, Sp3 can function as either an activator or repressor depending on the promoter and the cellular context (46,53). In the case of the ␤4 promoter, both Sp family members function as transcriptional activators (Refs. 30 and 31 and this report) and can do so in at least an additive, if not a synergistic, manner. Thus, the physiological relevance of hnRNP K totally repressing Sp1 function but not that of Sp3 remains obscure but may be related to one of the many potential functions of hnRNP K in regulating ␤4 gene expression (see below). Alternatively, it may be coupled to the appearance of the various factors during development. Although the expression patterns of Sp1, Sp3, and hnRNP K have not been studied extensively in terms of neuron-specific expression, other than a report describing high levels of Sp1 expression in embryonic neural tissue of mouse (54), it is possible that a developmental gradient of expression of the three factors exists such that Sp1 and hnRNP K are expressed early in development, and Sp3 is activated later. Interestingly, it has been demonstrated recently that cell cycle withdrawal and subsequent neuronal differentiation of the PC12 pheochromocytoma cell line, as induced by nerve growth factor, is at least in part caused by transcriptional activation by Sp1 of the genes encoding the cell cycle control proteins, the cyclin-dependent kinase inhibitor, p21 WAF1/CIP1, and cyclin D1 (55). Therefore, it is plausible that during early embryonic development, Sp1-induced expression of cyclin-dependent kinase inhibitors and cyclin D1 may lead to premature exit of specific neuronal precursors from the cell cycle. If hnRNP K is present in sufficient quantities at the same time in development as Sp1, it may prevent Sp1 from binding to its recognition sites in neuron-specific genes, particularly those encoding proteins involved in cell cycle withdrawal, thus allowing, at some level at least, normal development to proceed.
hnRNP K exhibits a broad array of regulatory functions leading to its characterization as a "nucleic acid-regulated docking platform" (50). As mentioned earlier, it interacts with RNA as well as single-and double-stranded DNA, with singlestranded DNA being the preferred substrate (50,56). This latter property has led to the suggestion that hnRNP K can facilitate transcription indirectly by functioning as an "architectural" transcription factor (57). This would be done by hnRNP K creating a single-stranded bubble that would make the DNA more flexible and thus allow other DNA-binding proteins to align more easily with the basal transcriptional machinery in a manner that might otherwise be energetically unfavorable (for a complete discussion of this potential function of hnRNP K, see Ref. 57). Such a model has been proposed for the regulation of the c-myc promoter by hnRNP K (58). Michelotti et al. (58) demonstrated that there are regions of singlestranded DNA in the c-myc promoter in cells that actively transcribe the c-myc gene, whereas such regions are absent in cells that do not express c-myc (58). One of these singlestranded DNA regions, the CT element, has been shown to be a binding site for hnRNP K (59). Indeed, hnRNP K was demonstrated to induce single-stranded DNA conformation at this site; moreover, opening of the CT element by hnRNP K was augmented in the presence of the transcriptional suppressor of the sterol regulatory element, CNBP, a single-stranded DNAbinding protein that interacts with the purine-rich strand of the CT region (58). Interestingly, in the context of the ␤4 promoter, the favored substrate for Pur␣, which also prefers binding to single-stranded DNA, is the lower strand of E1, thus FIG. 7. Western blot analysis of Sp1, Sp3, and hnRNP K in transfected Drosophila SL2 cells. 10 g of nuclear extracts prepared from SL2 cells transfected with expression constructs for the indicated proteins were analyzed by Western blotting with anti-Sp1, anti-Sp3, and anti-hnRNP K antibodies. The cells were transfected as described in Fig. 6. it is easy to imagine a bubble structure created by the binding of hnRNP K to the upper strand and Pur␣ to the lower strand. How this would lead to differential regulation of Sp1 and Sp3 is currently being investigated using all four proteins in structural and functional studies. In the case of the c-myc promoter, in addition to hnRNP K and CNBP, the CT element can also be bound by Sp1. However, in contrast to the ␤4 promoter, Sp1 had a dominant effect over hnRNP K and CNBP, eliminating a "bubble" they created on the c-myc promoter (58). Together these observations reflect further the potential complexity of functional interactions between hnRNP K and Sp1 family members, indicating that these interactions are dependent on cellular and promoter contexts.
In addition to nucleic acid interactions, hnRNP K binds to a variety of regulatory proteins, some of which are involved in signal transduction pathways (such as Src, Fyn, Lyn, and Vav; for review see Ref. 50), whereas others participate in transcriptional processes. The latter group of proteins include several that are particularly relevant to the present study. One of them, TATA-binding protein (TBP), can be coimmunoprecipitated with hnRNP K from nuclear extracts (42). This is of interest because the ␤4 gene promoter does not contain a TATA sequence, and therefore the coupling of the basal transcriptional machinery to the ␤4 promoter must occur independently of TBP binding to DNA. Protein-protein interactions (e.g. hnRNP K-TBP) may be one way by which this is achieved. Interestingly, Sp1 and Sp3 have also been postulated to tether the basal transcriptional machinery to TATA-less promoters via protein-protein with TBP-associated factors (60,61). Thus, three proteins we have shown to bind to regulatory regions of the TATA-less ␤4 promoter all have the potential to interact with the general transcriptional apparatus. We are currently studying the significance of this observation. Investigation of whether the remaining E1-binding proteins, NARP31 and NARP114, are relevant to the transcriptional regulation of the ␤4 gene awaits their identification.
In summary, we have identified hnRNP K as a regulatory factor that interacts with the promoter region of the neuronal nACh receptor ␤4 subunit gene. Furthermore, we have shown that hnRNP K regulates the transcriptional effects of Sp1 family members on ␤4 gene expression differentially. The repressive effect of hnRNP K on Sp1 transactivation represents a novel property of this multifunctional protein. The physiological significance of this phenomenon in the context of the ␤4 gene has yet to be determined; however, given the important roles hnRNP K plays in mRNA processing as well as transcription, we believe the present report provides further evidence that these two critical biological processes are intimately linked.