Multiple Forms of the Human Gene-specific Transcription Factor USF

VOl. 263. No. 24, Issue of August 25, PP. 11985-11993.1988 Printed in U.S.A.

THEJOURNALOF BIOLOGICAL CHEMISTRY 0 1988 by The American Society for Biochemistry and Molecular Biology, Inc.

Multiple Forms ofthe Human Gene-specific Transcription FactorUSF I. COMPLETEPURIFICATION

AND IDENTIFICATION OF USF FROM HeLa CELL NUCLEI* (Received for publication, January 14, 1988)

Michele Sawadogo$, Michael W. Van Dykeg, Polly D. Gregorll, and Robert G. Roeder From the Laboratory of Biochemistry and Molecular Biology, The Rockefeller Uniuersity, New York, New York 10021-6999

The human gene-specific upstream stimulatory transcription factor (USF) is required, both in vivo and in vitro, for maximal expression of the major late promoter (MLP) of adenovirus. We report here the complete purification and identification of USF from HeLa cell nuclei. The protein was followed throughout its purification using a quantitative filter binding assay. With a combination of classical purification techniques and fast-flow protein liquid chromatography, USF can be purified to homogeneity startingeitherwith a standard HeLa cell nuclear extract or with a higher salt extract from (lysed) HeLa cell nuclei (nuclear pellet extract). Approximately 20,000-fold purification from the nuclear pellet extract and 80,000-fold from the nuclear extract are necessary to obtain homogeneous preparations of the transcription factor. A maximum of 20,000 molecules of USF appear to bepresent in HeLa cells. Twomajor forms of the USF protein can be distinguished both by their slightly differentmobilities in sodium dodecyl sulfate gel electrophoresis (apparent molecular weights 44,000 and 43,000, respectively) and by different electrophoretic mobilities of the corresponding protein-DNA complexes. Both forms of USF are heat-stable and interact with the MLP as monomers. Antibodies elicited against purified HeLa USF interact with the transcription factor bound to the MLP upstream element.

The molecular mechanism leading to specific transcription initiation by mammalian RNA polymerase I1 is still very poorly understood, although it is clear that high transcription rates correlate with the participation of several DNA regulatory elements. A basal level of transcription can be observed, both in uiuo and in uitro, when all but a small region of the promoter DNA has been deleted. This minimum promoter is usually centered on a TATA box sequence (1). For this basal level of gene expression, at least three general transcription factors are necessary in addition to the RNA polymerase I1 (2-4), and thereis a requirement for ATP or dATPhydrolysis (5, 6). Footprinting analyses have demonstrated a direct interaction between one of the basic factors, designated RNA

* This work was supported inpart by National Institutes of Health Grants GM38212 (to M. S.) and CA 42567 (to R. G. R.) and by general support from the Pew Trust to the Rockefeller University. The costs of publication of this article were defrayed in part by the payment of page charges. This articlemust therefore be hereby marked “aduertisernent” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. $ Leukemia Society of America Special Fellow. I Supported by National Institutes of Health Postdoctoral Fellowship GM 09949. 7 Supported by National Institutes of Health TrainingGrant A107233.

polymerase I1 transcription factor D, and theTATA sequence (7, 8).’ The specific roles of the other general factors (RNA polymerase I1 transcription factors B and E) are as yet unknown. The highly phosphorylated form of RNA polymerase I1 (form 110) is required for specific transcription (9,lO). TWO classes of cis-acting DNA elements are involved in modulating the TATA box-driven transcription of specific genes. The socalled “upstream elements” act at short distance, whereas enhancer elements can be located up to several thousand base pairs upstream or downstream of the transcription initiation site (11, 12). So far, there is no definitive information on the mechanism by which these regulatory DNA elements (and their cognate transcription factors) control the rate of transcription initiation in eukaryotes. The major late promoter (MLP)’of adenovirus has proven to be a useful model system for in vitro analysis of specific transcription by RNA polymerase 11. The structure of this promoter appears relatively simple, with only two important elements: a TATA box at positions -31 to -25 and a unique upstream sequence at positions -63 to -52 (13-16). In the viral context, the MLP upstream sequence does not seem necessary for the early (Ela-dependent) expression of the promoter, but is absolutely required for late function (17). A HeLa nuclear protein, designated USF (or major late transcription factor or upstream element factor), has been shown to interact with the MLPupstream sequence and tostimulate transcription from this promoter in uitro (8, 18, 19). This gene-specific transcription factor appears to be a monomeric polypeptide with a molecular weight of 46,000-55,000 (20-23) which is involved in expression of cellular genes like rat yfibrinogen, metallothionein I, and possibly human kininogen (24-26). An analysis by scanning transmission electron microscopy was used recently to present a model for the threedimensional structure of USF bound to the MLP(22). In this paper, we describe the complete purification of USF from HeLa cell nuclei and present evidence for the existence of multiple forms of the protein. A detailed analysis of the properties of the purified transcription factor is presented in the accompanying paper (27). MATERIALS AND METHODS AND RESULTS3

Determination of USF Concentrations by Filter BindingUSF can be assayed on the basis of its stimulatory effect on N. Nakajima and R. G. Roeder, unpublished data. ‘The abbreviations used are: MLP, major late promoter; USF, upstreamstimulatory factor; SDS, sodium dodecyl sulfate; NPE, nuclear pellet extract; PMSF, phenylmethylsulfonyl fluoride; BSA, bovine serum albumin; PEG, polyethylene glycol. Portions of this paper (including “Materials and Methods,” part of “Results,” Figs. 1 and 7, Tables I, 111, and Table IV and Equations 1 and 2) are presented in miniprint at theend of this paper. Miniprint is easily read with the aid of a standard magnifying glass. Full size photocopies are included in the microfilm edition of the Journal that is available from Waverly Press.

11985

11986

Purification of USF from HeLa Cells

in vitro transcription from the adenovirus MLP (8).However, such an assay presents two disadvantages. First, it requires large amounts of complementary transcription factors which need to be free of contaminating USF,therefore necessitating purification through several chromatographic steps. Second, the assay is very sensitive to a variety of inhibitors and thus would not be quantitative with crude fractions. We therefore decided to take advantage of the specific DNA binding properties of USF in order to assay it independently of the other transcription factors. Filter binding assays rely on the property of nitrocellulose to retain proteins and protein-DNA complexes while allowing naked double-stranded DNAs to flow through. Using a small labeled DNA fragment containing a USF-binding site and a vast excess of cold carrier DNA to titrate out nonspecific DNA-binding proteins, it is possible to detect the presence of USF even in a complex mixture of proteins (8). Given the quantitative nature of a filter binding assay, this allows calculation of the molar concentration of USF in any given fraction (and therefore, with an estimate of the protein molecular weight, its degree of purity). We first established standard assay conditions which optimize for USF-specific binding and minimize nonspecific interactions of other proteins to the labeled DNA probe (see “Materials and Methods”). We then determined the apparentdissociation constant (&app)) of USF for its specific binding site under these particular conditions (Fig. 1).Given this &(epp) (7.5 X 10”’ M under our standard assay conditions) and the molarity of specific binding sites in the reaction known, one can calculate the total concentration of USF present in any reaction by measuring the concentration of specific complexes formed at equilibrium (see Equation 2 under “Materials and Methods”). One potential problem with this assay is that, with crude fractions, nonspecific interaction of abundant DNA-binding proteins with the radioactive probe may not benegligible despite the vast excess of cold competitor DNA present inthe reaction. To eliminate this uncertainty, two parallel reactions were routinely assembled one with 1 pg of p(C2AT)19 (4) as cold carrier DNA and the other with 1 wg of the MLPcontaining plasmid pML(C2AT)19. In the second reaction, the 80-fold molar excess of cold MLP effectively competed for USF binding to the labeled DNA fragment. This competition allowed for a measurement of the nonspecific background, which could then be subtracted for accurate calculation of USF concentrations. Extraction of USF from HeLa Cell Nuclei-USF was first detected in HeLa cell nuclear extracts prepared according to Dignam et al. (28). However, upon quantitation, the concentration of USF present in this kind of extract was found to be somewhat variable. This prompted us to investigate the conditions for maximizing the extraction of USF from HeLa cell nuclei. In the procedure of Dignam et al., the final salt concentration during the nuclei extraction is usually around 0.15-0.2 M. We found that theamount of USF extracted from the nuclei actually increases almost linearly as a function of the saltin the 0.15-0.3 M range (see Table I). At the optimum, the USF extracted corresponds to about 20,000 molecules/ cell. Although more USF can be extracted from the cells by using a higher salt concentration during the nuclear extraction, thespecific activity of the transcription factor in higher salt nuclear extracts is not significantly higher than that in lower salt nuclear extracts because of a parallel increase in the overall protein concentration. However, crude extracts with about 3-4-fold higher USF specific activity can be obtained by high salt extraction of the residual nuclei after the

extraction procedure of Dignam et al. (28) (nuclear pellet extracts). Preparationof these extracts resultsin nuclear lysis and therefore necessitates a high speed centrifugation of the chromatin (see “Materialsand Methods” for experimental details). Nuclear pellet extracts are a good starting material for USF purification. However, theyaretranscriptionally inactive, presumably due to thepresence of inhibitors, including histones or other tightDNA-binding proteins. Purification of USF from Nuclear Pellet Extracts-We will describe here in detail the purification of USF from nuclear pellet extracts. However, a very similar procedure can be used to obtain purified USF from standard nuclear extracts (see “Materials and Methods” and Table 111). A summary of two independent preparations of USF is presented in Table 11. Because of the relatively low abundance of the USF protein, the purification required a somewhat large volume of starting material (typically 400-500 ml of nuclear pellet extract). The first two steps of the purification, heat treatment and ammonium sulfate fractionation, were designed to decrease rapidly both the total protein and the volume to be handled at the first chromatographic step. The rest of the procedure was optimized according to thefollowing criteria: rapidity, minimum number of manipulations, and maintenance of the highest possible protein concentration throughout the purification. (We found that USF is a very stable but somewhat sticky protein. Most losses during purification seem to be primarily due to absorption of the protein onto surfaces.) We described earlier the heat stability of the USF protein (8). We took advantage of this property to provide a rapid and efficient first purification step. A brief heat treatment of nuclear pellet extracts at 70 “Cresults in the precipitation of about 90% of the totalprotein. Assessment of the exact USF recovery during this step is difficult because the abundance of nonspecific DNA-binding proteins in the extract renders the USF quantitation before the heat treatment very imprecise. Some of theUSF could possibly be trapped in the precipitate of denatured proteins. However, Western blot analyses of crude extracts before and after heat treatment have indicated that the USF recovery is usually quantitative? The USFprotein is somewhat hydrophobic and precipitates at fairly low concentrations of ammonium sulfate. This property was used in the second step of the purification, where USF in the heat-treated extract was both concentrated and further purified by differential ammonium sulfate precipitation (0.88-1.43 M ammonium sulfate cut). Purification of USF by chromatography on hydroxylapatite was carried out as a two-step procedure. A first passage was performed in the presence of ammonium sulfate. Under these conditions, USF flows through the column, together with about one-third to one-half of the total protein. A second hydroxylapatite column was run in the absence of ammonium sulfate, and bound USF eluted by a linear gradient of potassium phosphate (Fig. 2). Although this last column does not result in a large increase of specific activity, it was found essential in removing specific proteins which would otherwise remain as contaminantsat theend of the purification. The last two steps of the USF purification take advantage of the rapidity and high resolution of chromatography by fastflow protein liquid chromatography. The two columns, an anion exchanger (Mono Q ) and a cationexchanger (Mono S), were run back to back and together provided up to 150-fold purification (see Fig. 2 and Table 11). Note that on the last column, the USF activity was reproducibly found to coelute

* P. D. Gregor, unpublished result.

Purification of USF HeLa from

Cells

11987

TABLE I1 Purification of USF from nuclear pellets extract Experimental details can be found under “Materials and Methods.” The numbers shown are the average for two independent purifications, each starting with nuclei isolated from about 3 X 10” HeLa cells. The USF concentrations were determined using the standard filter binding assay. Protein concentrations were measured by the Bradford assay (30) except for the Mono S fractions, where the amount of USF was determined by comparison with known amounts of bovine serum albumin on Coomassie Blue-stained gels. The numbers were then calculated assuming that USFin these fractions was greater than 70% pure. Purification step

I. 11. 111. IV. V. VI. VII.

Volume

Protein

USF

Specific activity

ml

mg

pmol

pmollmg

Yield

Purification -fold

2,532 Crude extract 489 11 1 1” Heat treatment 467 249 2,825 0.70 45 4 Ammonium sulfate ppt 9.2 43.9 1,965 1st hydroxylapatite 12.3 6 21.3 0.54 72 1,530 6.3 740 118 0.26 10 2nd hydroxylapatite 7.7 431 1,818 0.15 161 Mono Q 1.6 0.237 co.011 >16,727 184 >1,480 0.07 Mono S 1.5 a By assuming a quantitative recovery of USF during the heat treatment (see text), this stepprovides a 10-fold purification. Therefore, the overall purification from the crude extract to the Mono S fraction can be estimated to greater than 15,000-fold.

with one of the protein peaks detected by absorption at 280 nm (Fig. 2). Polypeptide Composition of Purified HeLa USF-Fractions at various steps of the USF purification were analyzed for their polypeptide content by electrophoresis on SDSgels (Fig. 3A). Coomassie Blue staining revealed that thevery complex composition of the first fractions simplified to a few polypeptides at theMono Q step and toonly two polypeptides at the Mono S step. By comparison with known protein markers, the upper (u) and lower ( I ) polypeptides in the final USF preparation hadmigrations corresponding to apparent molecular weights of 44,000 and 43,000, respectively (Fig. 3B). It is worth noting that when silver (rather than Coomassie Blue) was used for staining thegel (36), the 44-kDa polypeptide was much more difficult to detect, often staining with a pale yellow color, whereas the 43-kDa polypeptide stained rapidly as a dark brown band (data notshown). Interestingly, the relative abundance of the 44- versus 43-kDa proteins, as revealed by Coomassie Blue staining, was reproducibly a 1:2 ratio whenever the starting material had been nuclear pellet extract. When USF was purified from nuclear extract, thetwo protein bands were usually of comparable intensity (data notshown). Multiple Forms of USF Protein-The ability to quantitate precisely by filter binding assay the molar concentration of active USF in any given fraction provided us with a means to estimate the degree of purity obtained after the last column. As illustrated in Fig. 4, the 43- and 44-kDa polypeptides were found to coelute perfectly with the USF activity throughout the Mono S gradient. Comparison with known amounts of bovine serum albumin stained with Coomassie Blue under the same conditions proved unambiguously that at least one and most probably both of these polypeptides were active USF. Indeed, no other protein in the Mono S fractions was abundant enough to account for the activity determined by filter binding assay. In addition, the sum of the molar concentrations of the 43- and 44-kDa polypeptides correlated perfectly with the molar concentration of USF measured in each fraction. This was a very strong indication for USF being a monomer of 43 or 44 kDa, with all or nearly all of the purified protein being active (at least in terms of specific DNA recognition). Further evidence for the identity of both the 43- and 44kDa polypeptides as USF was provided by renaturation experiments (32). The two polypeptides were separated by electrophoresis in the presence of SDS, individually excised from

the stained gel, and then extractedand renatured asdescribed under “Materials andMethods.” Specific DNA binding to the MLP upstream sequence was monitored by DNase I footprinting (Fig. 5). Both renatured proteins showed cleavageprotection patterns which were indistinguishable from those previously described for USF (8). We could, however, observe a very slight but reproducible difference between the footprints of the two USF proteins: the intensity (but not the location) of the enhanced DNase I cleavages at the 5’ border of the footprint was always stronger for the 44-kDa form than for the 43-kDa form of USF (Fig. 5, compare lanes 4 and 5). Gel Shift Analysis of USF. DNA Complexes-Interestingly, the same heterogeneity observed for the purified USF protein by SDS gel electrophoresis was also detected when USF. MLP DNA complexes were analyzed by gel retardation assay (37, 38) (Fig. 6).This heterogeneity did notresult from some partial modification or proteolysis of the protein occurring during the purification since the exact same pattern observed with purified USF (Fig. 6A, lane 6) was also seen for USF present in crude nuclear extract (lane 2) or nuclear pellet extract (lane 4 ) . To establish clearly the relationship between the two major forms of USF .DNA complexes observed by gel retardation assay and the 43- and 44-kDa forms of purified USF, we used again the polypeptides renatured after separation on an SDS gel (see above and Fig. 5). The results of this experiment are shown in Fig.6B. The renatured 43-kDa protein (lane 3 ) gave a gel shift identical to thefaster migrating complex observed with the input material (lane I ) . The 44-kDa protein (lane 2 ) gave two shifted bands of equal intensity: one corresponding to the higher of the two major complexes in the input material and an additional slower migrating complex. This latest shift was also observed in the input material, but only as a very minor band. These results indicated that the various shifted bands observed for USF in gel retardation assays did indeed reflect the binding of different forms of the transcription factor, forms which could also be distinguished by their slightly different mobilities on SDS gel. However, it is not clear whether the highest shifted band observed most strikingly with renatured 44-kDa USF corresponds to an alternate conformation of this polypeptide or whether there is a third form of the USF protein which also migrates in SDS gel with an apparent molecular weight of 44,000 and renatures more efficiently than the other (major) 44-kDa form. It should be noted that the electrophoretic separation of these various USF. DNA complexes was highly


11988

A

Hydroxylapatite

B

I5

0

x

I lo " " "

U In

3

05

0 I

20

IO

30

50

40

70

60

; ,ol, ,

80

\,

Cytochrome C

2 x

MONO 0

0

3-

_"""

2 -

c

- 0.2

I -

a 0I

-

,

,

2

3

4

1

4

6

R E L A T I V E MOBILITY

I U In

1

,

'1

20

IO

30

1

40

50

60

'

1

0.10

MONO S

c

m

1

I

IO

20

30

40

50

60

FIG. 3. Monitoring course of USF purification by SDS gel electrophoresis.A, aliquots of various steps in the USFpurification (see Table 11) were analyzed for their polypeptide content by electrophoresis on SDS gel ( l a n e s 1-6), along with a mixture of protein molecular weight markers ( l a n e M @-galactosidase,phosphorylase b, serum albumin, ovalbumin, carbonic anhydrase, and cytochrome c). About 3 pg of protein were analyzed for each of the following: the nuclear pellet extract before ( l a n e 1 ) and after( l a n e 2) heat treatment, the ammonium sulfate precipitate ( l a n e 3), and thefirst ( l a n e 4 ) and second ( l a n e 5) hydroxylapatite fractions. For the Mono Q ( l u n e 6 ) and Mono S ( l a n e 7) fractions, the aliquots loaded on the gel contained 0.8 and 0.1 pg of total protein, respectively. B, the relative mobilities on SDS gel of the upper (u)and lowerr ( I ) USF polypeptides (Mono S fraction) were determined by comparison with the migration of protein molecular weight markers. The semilogarithmic plot of these results indicated apparent molecular weights of 44,000 and 43,000, respectively, for the two USF polypeptides.

70

FroctiN o nu m b e r

FIG. 2. Column chromatography for purification of USF. From top to bottom, the three hydroxylapatite, Mono Q, and Mono S columns represent the last successive steps in the USF purification. were determined by filter In each case, the USF concentrations (e) binding assay as described under "Materials and Methods." Upper, hydroxylapatite column. The protein concentrations (0)were determined by the method of Bradford (30). Fractions were 1 ml each. Center, Mono Q column. Shown is the actual recording by the UV monitor a t 280 nm. 0.5-ml fractions were collected. The protein profile on this column was somewhat variable from preparation to preparation. However, thesalt elution of the USF peak was absolutely reproducible, with about 80% of the activity contained in three to four fractions between 185 and 205 mM salt. Lower, Mono S column. The actual recording of the UV monitor is shown. The fractions (3 drops, 125 pl) were collected in siliconized Eppendorff tubes containing 75 pl of sterile glycerol and 1 pl of 1 M dithiothreitol. The USF concentrations were determined after careful mixing of the protein frations with the additional glycerol.

dependent upon the length of the DNA fragment. Larger fragments (>200 base pairs) allowed for the best resolution, whereas all the complexes with short oligonucleotides (30 base pairs) showed nearly identical mobilities (data not shown). General Properties of Purified HeLa USF-Purified USF was perfectly stable for at least 1 year at -20 "C in the presence of 50% glycerol and 5 mM dithiothreitol. However, care had to be taken because of the very low protein concentration, and pipetting was done only with siliconized plastic tips (until dilution in the presence of carrier bovine serum albumin and detergent). The heat stability of purified USF was comparable to that of cruder fractions (Table IV). Transcriptional activity, however, was significantly lower (see accompanying paper (27)). No ATPase, DNase, topoisomerase, or protein kinase activity

" u

*

-

"

-

Il697 66 I

45

L

29

X

12

FIG.4. Coelution of 43- and 44-kDa polypeptides with USF on Mono S column. The polypeptides present in various fractions from the Mono S column shown in Fig. 2 were resolved by SDS gel electrophoresis and stained with Coomassie Blue. In each case, 8 pl of the fraction were loaded in the indicated lanes. Ft refers to the 2, lower). The flow-through material (pooled fractions 4-26;Fig. molecular weight markers were as described for Fig. 3. The picomoles of USF applied to each lane were calculated from the concentrations determined by filter binding assay. Visual comparison with known amounts of bovine serum albuminapplied to anidentical gel indicated that the43-kDa polypeptide in the 8 pl of fractions 62-67 was present at about 20, 40, 60, 60, 40, and 20 ng, respectively. The 44-kDa polypeptide appeared about half as abundantin each case.

could be detected in the purified USF preparations (data not shown). Interaction of Specific Mouse Antibodies with USF.DNA Complexes-Polyclonal antibodies against USF were raised in a mouse as described under "Materials and Methods.'' In Western blots, the serum reacted with the 43-kDa protein previously identified as USF and to a lesser extent with the 44-kDa protein (Fig. 7). To provide evidence for specific binding of antibodies to USF. DNA complexes,a gel retarda8). Preformed USF. DNA complexes tion assay was used (Fig. were incubated for 25 min at room temperature with serial dilutions of immune or naive mouse serum and thenseparated by electrophoresis. As increasing concentrations of immune

Purification of BOTTOM S TSRTARNADN D

USF from HeLa Cells

TOP

11989

A

B

“ G I 2 3 4 5 0

Purified

I 2 3 4 5

NE NPE USF

Specific competitor

- 3 ”=

I 2 3 4 5 6 7 e ” “ ”

FIG.6. Gel retardation assay of USF-MLP DNA complexes. A, comparison between crude and purified USF fractions. Binding reactions (20 p l ) contained 3.9 pg of [d(I.C)]. and1.7 pg of p(CzAT)19 (4) as cold carrier DNA, 0.4 ng of a 32Pend-labeled 260-base pair DNA fragment encompassing MLP, and either 40 ng of the USFspecific oligonucleotide (see “Materials and Methods”) where indiFIG. 5. DNase I footprints of renatured 44- and 43-kDa cated or 40 ng of a nonspecific oligonucleotide of the exact same size. USF polypeptides on MLP. The two USF polypeptides were re- Protein addition was as follows: none (lane I ) , 0.17 pl of nuclear natured afterisolation from an SDSgel as described under “Materials extract (NE; lanes 2 and 3), 0.25 pl of nuclear pellet extract (NPE; and Methods.” Footprinting reactions were carried out as described lanes 4 and 5),or 0.1 ng of purified USF (Mono S fraction; lanes 6 earlier (8).The top strand was 5’ end-labeled, and thebottom strand and 7).After a 30-min incubation a t 30 “C,the samples were processed was 3’ end-labeled. In both cases, the reactions were run in parallel as described under “Materials and Methods.” The migrations of the with a G-specific sequencing ladder of the same fragment (39). For free DNA probe (Unbound)and of the upper (u)and lower (1) USFeach DNA strand, lanes 1 and 3 are control reactions. The footprints containing complexes are indicated. B, interaction of the gel-purified obtained with purified USF (Mono S fraction) is shown in lane 2. USF polypeptides with MLP DNA. Binding reactions (10 p l ) were Lanes 4 and 5 show the footprints obtained with the renatured 44- assembled using as a probe, 0.8 ng of a “P end-labeled 450-base pair and 43-kDa polypeptides, respectively. MLP-containing DNA fragment. For the control lane, 0.06 ng of purified USF (Mono S fraction) was used. Equivalent amounts (in serum were added, a series of shifted bands of lower mobility terms of DNA binding activity) of the gel-purified 44- and 43-kDa were seen (lanes 2-7), and the complexes normally seen in USF polypeptides were used in the other reactions as indicated. The the absence of added serum decreased in intensity. In the two major USF.MLP DNA complexes in the control reaction are indicated as u (upper band) and 1 (lower band). A higher shift (u’)is presence of undiluted immune serum, the USF .DNA com- mostly observed with the renatured 44-kDa protein (lane 2), but can plexes werealmost entirely replaced byone or more complexes be seen also as a minor band in the control lane.

of much lower mobility( l a n e 7). In contrast,addition of serum from a naivemouseshowed no effect on the USF. DNA complexes (lanes 8-10),except for additional bands of higher mobility seen when undiluted naive serum was added ( l a n e 10). We attribute theseshifted bands to proteases present in the serum. From these results, we concluded that the immune serum contains antibodies which specifically bind USF-DNA complexes. The multiple species of shifted bands suggest that USF contains several antigenic determinants; and that, at high concentrations of serum, all are occupied by specific antibodies. At lower serum concentrations, on average, only one determinant is complexed to an immunoglobulin molecule.

DNA complexes is very similar for specific and nonspecific sites: Thus, we think that the procedure would not be as efficient in the case of USF, although other strategies using inclusion of carrier DNA during binding to theaffinity resin may be applicable (20, 41, 45). As an alternative, however, our present purification scheme combines a series of steps which each separates proteins on totally different principles: heat stability, solubility in the presence of ammonium sulfate (which is somewhat related to the hydrophobicity of the protein), interaction with hydroxylapatite (whose chemistry of interaction is not yet totally understood), and, finally, separation according to the distribution of charges at the protein surface on both types of ion exchangers. The use of fast-flow protein liquid chromatography columns allowed for DISCUSSION a resolution in these last two steps that classical chromatogWe have purified and identified unambiguously the gene- raphy resins would not offer. Our final USF recovery was specific transcription factor (USF) from HeLa cells. Because somewhat low (usually between 4 and 10%). As mentioned the protein is not very abundant, ahigh degreeof purification earlier, most losses during the USF purification seem to be was required to reach homogeneity (80,000-fold purification due to absorption of the protein on surfaces. This problem necessary starting from nuclear extract). In several instances, could potentially be eliminated by including nonionic deteraffinity matrices have been used successfully to purify se- gents in the purification buffers. quence-specific DNA-bindingproteins (40-42). For some proWe found that purified USF is perfectly stable to temperteins, specific DNA binding is much more resistant to salt atures as high as 100 “C.This could indicate that the active than is nonspecific DNA binding. In those cases, successive conformation of the protein is also thermodynamically its chromatography on nonspecific DNA columns followedby most stable conformation. Others have reported a !ackof specific DNA columns can provide a high degree of purificaM. Sawadogo, unpublished result. tion (40, 43, 44). By contrast, the salt stability of the USF.


11990 Immune serum

Naive E m Serumdilution

I2345678910

1

Ig-USF-DNA

3 USF-DNA

J FreeDNA FIG.8. Interaction of USF-specificantibodies with proteinDNA complexes. USF.DNA complexes were formed using the specific oligonucleotide as described under “Materials and Methods.” 10 pl of preformed complexes were aliquoted into tubes containing 1 pl of the indicated serum dilutions. Following a 25-min incubation a t room temperature, the samples were processed as described under “Materials and Methods.” The control lane containedthe USF-DNA complexes diluted with serum dilution buffer alone. The location of the free probe, the USF-DNA complexes, and the immunoglobulin (Ig).USF.DNA complexes are shown. The arrow indicates the location of those USF-DNA complexes containing the intact 44- or 43kDa protein. The minor shifted band indicated with a star reflects binding of a degradation product of USF which is also recognized by the antibodies.

stability for USF (major late transcription factor)at temperatures higher than 50 “C (20). The most likely explanation for this discrepancy relates to the hydrophobicity of the transcription factor: at high temperatures, the protein could be lost by nonspecific adsorption onto surfaces unless carrier protein and/or detergents are present. We demonstrated the existence of both 44- and 43,000-Da forms of USF in preparations derived from HeLa cells. Whether this observation bears any physiological significance remains to be investigated. The two proteins have clearly identical DNA binding properties (27) as well as similar chromatographic behavior. However,their very different electrophoretic mobility in polyacrylamide gels when bound to several hundred-base pair-long DNA fragments distinguishes them to anextent which couldnot be predicted from the very slight difference in their molecular weights. Wedo not know yet the extent of the immunological relationship between the two proteins. Our polyclonal mouseserum reacts better with the 43-kDa protein than with the 44-kDa protein. This was observed both on Western blots (Fig. 7) andby gelretardation assay using the renatured gel-purified USF polypeptides (data not shown). However, we could demonstrate the presence of antibodies in this serum recognizing several distinct determinants on both USF. DNA complexes (Fig.8). It is possible that the smaller (43 kDa) form of USF would be generated artificially in vitro from the larger (44 kDa) polypeptide by, for instance, limited proteolysis or dephosphorylation. If this is the case, this modification must occur during the extraction procedure since both forms of the protein can be seen already in the crude extracts (Fig. 6). Alternatively, these various forms may be present in the cell. Oce would then have to determine whether these reflect various modifications of the

same protein, alternate splicing mechanisms, or the existence of different genes. Such questions will be better addressed once a USF cDNA clone is obtained. With the availability of homogeneous preparations of USF, we are now able to pursue a number of studies which would otherwise be impossible given less purified material. These include determination of partial amino acid sequence for cDNA cloning, detailed investigation of the USF/DNA interactions, and preparationof specific antibody probes for studying structure-function relationship and investigating possible post-translational modifications. Hopefully, these studies will bring new insights into the mechanism of action of USF in particular and of upstream transcription factors in general. We will report in the accompanying paper (27) some of the DNA binding and transcriptional properties observed with purified HeLa USF. Acknowledgments-We wish to thank Carmen-Gloria Balmaceda and Marie-Caroline Weichs an der Glon for excellent technical assistance,Elizabeth Slattery for the aliquots of extracts made a t various salt concentrations, and Nathaniel Heintz for synthesizing the USFoligonucleotide. REFERENCES 1. Breathnach, R., and Chambon, P. A. (1981) Annu. Rev. Biochem. 50,349-383 2. Matsui, T., Segall, J., Weil, P. A., and Roeder, R. G. (1980) J. Biol. Chem. 255,11992-11996 3. Samuels, M., Fire, A., and Sharp, P. A. (1982) J. BWl. Chem. 257,14419-14427 4. Sawadogo, M., and Roeder, R. G. (1985) Proc. NatL Acad. Sci. U. S. A. 82,4394-4398 5. Bunick, D., Zandomeni, R., Ackerman, S., and Weinmann, R. (1982) Cell 29,877-886 6. Sawadogo, M., and Roeder, R.G. (1984) J. BWL Chem. 2 5 9 , 5321-5326 7. Parker, C. S., and Topol, J. (1984) Cell 36,357-369 8. Sawadogo, M., and Roeder, R. G. (1985) Cell 43,165-175 9. Bartholomew, B., Dahmus, M.E., and Meares, C. F. (1986) J. Biol. Chem. 261,14226-14231 10. Cadena, W. S., and Dahmus, M. E. (1987) J. BWl. Chem. 262, 12468-12474 11. Dynan, W. S., and Tjian, R. (1985) Nature 316,774-778 12. Khoury, G., and Gruss, P. (1983) Cell 33,313-314 13. Hu, S. L.,and Manley, J. L. (1981) Proc. Natl. Acad. Sci. U.S. A. 78,820-824 14. Hen, R., Sassone-Corsi, P., Corden, J., Gaub, M. P., and Chambon, P. A. (1982) Proc. Natl. Acad. Sci. U. s. A. 79,7132-7136 15. Concino, M., Goldman, R. A., Caruthers, M. H., and Weinmann, R. (1983) J. Biol. Chem. 258,8493-8496 16. Miyamoto, N. G., Moncollin, V., Wintzerith, M., Hen, R., Egly, J. M., and Chambon, P. A. (1984) Nucleic Acids Res. 12,93099321 17. Logan, J., and Shenk, T. (1986) Nucleic Acids Res. 14, 63276335 18. Carthew, R. W., Chodosh, L. A., and Sharp,P. A. (1985) Cell 4 3 , 439-448 19. Miyamoto, N. G., Moncollin, V., Egly, J. M., and Chambon, P. (1985) EMBO J. 4,3563-3570 20. Chodosh, L. A., Carthew, R. W., and Sharp, P. A. (1986) Mol. Cell. BWl. 6,4723-4733 21. Moncollin, V., Miyamoto, N. G., Zheng, X. M., and Egly, J. M. (1986) EMBO J. 5,2577-2584 22. Hough, P. V. C., Mastrangelo, I. A., Wall, J. S., Hainfeld, J. F., Sawadogo, M., and Roeder, R. G. (1987) Proc. Natl. Acad. Sci. U. S. A. 84,4826-4830 23. Lennard, A. C., and Egly, J. M. (1987) EMBO J. 6,3027-3034 24. Chodosh, L. A., Carthew, R. W., Morgan, J. G., Crabtree, G. R., and Sharp,A. P. (1987) Science 238,684-688 25. Carthew, R. W., Chodosh, L. A., and Sharp, P.A. (1987) Genes Deu. 1,973-980 26. Sawadogo, M., and Roeder, R. C. (1986) in Cancer Cells/DNA

Tumor Viruses (Grodzicker, T., Sharp, P., and Botchan, M.,

Purification of USF from HeLa Cells 27. 28. 29. 30. 31. 32. 33. 34. 35. 36.

eds) Vol. 4, pp. 147-154, Cold Spring Harbor Laboratory, Cold Spring Harbor, NY Sawadogo, M. (1988) J. Biol. Chern. 263,11994-12001 Dignam, J. D., Lebowitz, R. M., and Roeder, R. G. (1983) Nucleic Acids Res. 11, 1475-1489 Wood, W. I. (1976) Anal. Biochern. 73, 250-257 Bradford, M. M. (1976) Anal. Biochern. 72, 248-254 Laemmli, U.K. (1970) Nature 227,680-685 Hager, D. A., and Burgess, R. R. (1980)Anal. Biochern. 109, 7686 Maniatis, T., Fritsch, E. F., and Sambrook, J. (1982) Molecular Cloning: A Laboratory Manual, pp. 468-469, Cold Spring Harbor Laboratory, Cold Spring Harbor, NY Sakato, N., and Eisen, H. N. (1975) J. Exp. Med. 1 4 1 , 14111425 Towbin, H., Staehelin, T., and Gordon, J. (1979)Proc. Natl. Acad. Sci. U. S. A. 76,4350-4354 Oakley, B. R., Kirsch, D. R., and Morris, N. R. (1980) Anal. Biochem. 1 0 5 , 361-363

"SF

NUCLEI

11991

37. Garner, M. M., and Revzin, A. (1981) Nucleic Acids Res. 9,30473060 38. Fried, M., and Crothers, D. M. (1981) Nucleic Acids Res. 9,65056525 39. Maxam, A. M., and Gilbert, W. (1980) Methods Enzymol. 65, 499-560 40. Rosenfeld, P. J., and Kelly, T. J. (1986) J . Biol. Chern. 261, 1398-1408 41. Kadonaga, J. T., and Tjian, R. (1986) Proc. Natl. Acad. Sci. U. S. A . 83,5889-5893 42. Wu, C., Wilson, S., Walker, B., Dawid, I., Paisley, T., Zimarino, V., and Ueda, H. (1987) Science 2 3 8 , 1247-1253 43. Treisman, R. (1987) EMBO J. 6, 2711-2717 44. Prywes, R., and Roeder, R.G. (1987) Mol. Cell. Bwl. 7, 34823489 45. Levens, D., and Howley, P. M. (1985) Mol. Cell. Biol. 5, 23072315

11992


0.8 0.6

0.4

0.2 0

I/ 111

I" \' "I


11993

................................................................... llr.l:-:,.('