Expression in Escherichia coli and Characterization of the Heat-stable Inhibitor of the CAMP-dependent Protein Kinase*

Pure heat-stable inhibitor of the CAMP-dependent protein kinase (PKI) has been isolated in high yield by using a bacterial expression vector constructed to syn-thesize the complete sequence of the rabbit muscle protein kinase inhibitor, plus an amino-terminal initiator methionine and glycine. Bacterially expressed PKI has an inhibitory activity identical to that of the protein isolated from rabbit skeletal muscle and, by gel filtration and gel electrophoresis, has the same physi-cochemical characteristics as the native physiological form of PKI. Fourier transformed infrared spectros-copy and CD establish that PKI has unusually large amounts of random coil and turn structures, with sig-nificantly smaller amounts of a-helix and B structures. The heat-stable inhibitor of the CAMP-dependent protein kinase is a small protein which binds to the catalytic subunit of the kinase and inhibits its activity 2). The effects of PKI’ are highly selective for the CAMP-dependent kinase; PKI does not inhibit any other serine/threonine kinase, even the closely homologous cGMP-dependent protein kinase (1, 3). While the physiological function of PKI is not yet apparent, it is clear that PKI is exceptionally potent with a Ki in

subunit, it would be desirable to obtain large amounts of the protein. While it has been possible to purify PKI from natural sources, the very low concentration of the protein has limited the availability of pure protein. Similarly, while synthetic peptides representing portions of PKI have been extremely useful, unlike intact PKI, these peptides inhibit kinases other than the CAMP-dependent protein kinase at high concentrations (3). Therefore, the complete PKI, rather than synthetic peptides, would likely be the most useful for studies addressing selective inhibition of kinase activity. In the present study, a bacterial expression vector for rabbit muscle PKI has been prepared based upon a synthetic gene that we have previously synthesized (12). PKI has been purified to apparent homogeneity in large amounts from bacteria carrying the expression vector. This has permitted the exploration of the properties of PKI that were not previously possible because of the limited amounts of material available.

RESULTS
Bacterial Expression of PKI-The bacterial expression vector for the rabbit muscle PKI was constructed using a synthetic PKI coding sequence which was previously used to prepare a mammalian expression vector (15). Essentially, this involved conversion of an NcoI site which contained the initiation codon to an NdeI site ( Fig. 1) and insertion of the fragment into pT7-7 a derivative of the 7'7-1 expression vector (14). This expression vector contains a T7 RNA polymerase promoter upstream of a ribosome binding site and polylinker. The NdeI site of the vector polylinker contains an ATG which is appropriately placed relative to the ribosome binding site to facilitate translation initiation. Therefore, conversion of the NcoI of the PKI coding sequence to an NdeI site should permit high level translation of the PKI sequence. It should be mentioned that the mammalian PKI expression vector encoded both an initiator methionine and additional glycine codon at the amino terminus of PKI. The glycine was added to the PKI coding sequence to produce a protein with favorable translation initiation properties for expression in mammalian cells (22). The NcoI to NdeI conversion for preparation of the bacterial expression vector should maintain the presence of the methionine initiation codon and the additional glycine codon. The correct construction of the T7-7-PKI clone was assessed by restriction enzyme digestion to confirm that the NdeI site had, in fact, been constructed. The T7-7-PKI DNA was then transfected into cells containing pGP1-2 in which the T7 RNA polymerase gene is controled by X PI..
Because the cells contain a temperature-sensitive X repressor, expression of the T7 RNA polymerase and subsequent expression of the PKI can be obtained by simply shifting the cells to the appropriate temperatures. Furthermore, because the T7 RNA polymerase is rifampicin-resistant, in contrast to the Escherichia coli enzyme, selective synthesis of PKI can be obtained by incubating the cells in rifampicin (14).
Bacteria carrying the T7-7-PKI expression vector were used to purify PKI. Analysis of crude extracts obtained from cells carrying the T7-7-PKI expression vector demonstrated the presence of material which inhibited the activity of the CAMPdependent protein kinase (Fig. 2). Low, but detectable inhibition of kinase activity was also obtained with extracts from cells which did not contain the PKI vector (data not shown). Thus, the assay of PKI activity in the crude extract may slightly overestimate the inhibitor content. Because the PKI is heat-stable, the initial step for purification of PKI involved heating the crude bacterial extract to 95 "C for 5 min. Following the heat treatment, gel electrophoresis demonstrated that the most abundant protein in the extract is a band with the appropriate mobility for PKI ( Fig. 3). At this stage of purification, extracts from bacterial cells which did not contain the PKI expression vector did not inhibit CAMP-dependent protein kinase activity and thus, after this step, analysis of preparations should accurately reflect the content of PKI.
PKI was further purified by ion exchange and gel exclusion chromatography. Again, the purification was monitored by analysis of PKI activity (Fig. 2) and SDS-polyacrylamide gel electrophoresis (Fig. 3). Most of the PKI activity which could be recovered was found to elute in the 100-300 mM fraction on DEAE-chromatography, and gel electrophoresis suggested the presence of only minor contaminants at this stage of the purification. The minor contaminants were readily removed by chromatography on Sephadex G-50 yielding an apparently homogeneous preparation. The results of a typical purification are summarized in Table I. Starting from 6 liters of bacterial culture, a final preparation of 14 mg of purified PKI was obtained with an apparent overall yield of 9%. While analysis of activity suggests that only a 12.8-fold purification was necessary, this is probably an underestimate due to the presence of nonspecific inhibitory activity present in the crude extracts. The recovery was good for most steps, except for the Sephadex G-50 columns. Somewhat low recovery was encountered at this step because of the need to utilize only the peak fractions to avoid contamination with other small proteins.
Characteristics of Bacterially Expressed PKI-The bacterially expressed PKI exhibited inhibitory potency identical to that of the native skeletal muscle protein. Both equally titrate the activity of pure CAMP-dependent protein kinase and exhibit identical specific activities (Fig. 4a). Henderson analysis (23) yielded K, values of 0.098 nM for the bacterial protein compared to 0.085 nM for the native protein; these values are identical within experimental error (Fig. 46). Multiple forms of PKI, designated I and 1', have been isolated from skeletal muscle (18). Of these only the I form is believed to be physiological (24). The bacterially expressed PKI exhibits characteristics identical to the physiological I form. It displays a molecular weight of 22,000 by gel exclusion chromatography (Fig. 5), quite distinct from the 11,000 apparent molecular weight characteristic of the I' form (18). Likewise, with Weber-Osborn gel electrophoresis (25), a method readily diagnostic for the two species, the bacterially expressed protein migrated identical to the I form (Fig. 6a). Both the bacterially expressed and native PKI migrated identically in either Weber-Osborn or Laemmli gel electrophoretic systems (Fig. 66).
In particular, the mixing experiment (Fig. 6b, lanes 4 and 7 ) demonstrates that the rabbit skeletal muscle and the recombinant PKI are indistinguishable by these two SDS-gel systems. From these combined experiments, it is clear that the bacterially expressed protein exhibits both activity and shape characteristics indistinguishable from that of the native protein. Structural Conformation of PKI from FTIR and CD- Fig.   7 shows the buffer-subtracted FTIR spectrum of PKI and its second derivative spectrum obtained after smoothing the raw data (over 9 points where the points are separated by 1. Absorbance spectra were measured for PKI in D20. Spectra were obtained after a 25-min purge with dry N, and were ratioed against a single-beam background collected with no cell. The spectrum of a buffer blank was subtracted from a sample spectrum to give spectra free of contributions from the buffer and cell. The lower trace is that of the second derivative (details of this procedure are described under "Experimental Procedures"). .. cm"). The deuterium-shifted amide region, amide 1', occurs between 1620 and 1700 cm" and primarily contains bands attributable to the carbonyl stretching vibrations contributed by each peptide linkage. On the basis of 21 proteins of known structure, Byler and Susi (26) have assigned 11 well-defined frequencies in the amide I' region to secondary structural elements. In the amide I' region the PKI spectrum shows a single broad peak whose maximum occurs at 1645 cm", the characteristic frequency assigned to random coil structure, indicating that this might be the dominant secondary structure in the protein. By taking the second derivative of the spectrum subtle variations in the line slope caused by contributions from multiple bands beneath the spectral envelope can be accentuated. The second derivative spectra indicate that there are bands at 1675,1663,1653,1645, and 1637 cm", and possibly at 1681 and 1632 cm". The 1675-1681-cm-' and 1632-1637-cm" bands can be assigned to the high and low components of extended chain structures; at 1663 cm" to turns or bends and at 1653 cm-I to a-helix. Resolution enhancement by Fourier deconvolution also shows evidence for a distinct band at each of these values, but the resolution of these bands is sufficiently poor that curve fitting, in order to estimate relative intensities and hence relative percentage secondary structure based on the Byler and Susi assignments, was not possible. However, it is clear from the maximum in the spectrum at 1645 cm" that a significant portion of the structure of PKI is likely to be random coil in nature.
One possible caveat to the interpretation of the FTIR data is that distorted helix structures have been observed to give rise to peaks as low as 1645 cm-' (27,28). In order to check this possibility, as well as to better quantitate the secondary structure, CD spectra were measured (Fig. 8). There is a deep minimum in this spectrum near 200 nm characteristic of a protein with a large proportion of random coil structure (29). The solid line is the CD spectrum calculated by the fitting procedure described under "Experimental Procedures." The calculated values for secondary structure in the protein based upon the fitted spectrum are 21 k 2% a-helix, 0 k 0% parallel 0-sheet, 15 +-4% antiparallel 0-sheet, 34 k 2% turn, and 30 f 2% other values including random coil. This result is in good agreement with the FTIR spectra. Thus both CD and FTIR support the conclusion that PKI contains a large proportion of random coil structure. The combined percentages of random coil and turn structures indicated by CD for this protein are unusually large.

DISCUSSION
Insertion of the coding sequence for rabbit muscle PKI in a bacterial expression vector has permitted the reproducible isolation of large amounts of PKI. In contrast to the isolation of the inhibitor from mammalian tissues, which requires multiple chromatographic separations and at least 10 kg of muscle tissue to isolate submilligram quantities (4), use of the bacterial expression vector permits relatively convenient and rapid isolation of 10-20-mg quantities of PKI. In several different preparations we have obtained at least 1 mg of PKI for each liter of bacterial culture. The inhibitory activity of bacterially expressed PKI is indistinguishable from that of the native skeletal muscle protein (Fig. 4).
Skeletal muscle PKI has been shown to adopt two conformations (18). One form, designated I, is the only form believed to be present physiologically (24). The other form, designated 1', appears most likely to be an artifact produced during extensive purification (4, 18, 24). The bacterially expressed PKI exhibits the charcteristics of the native I conformation (Figs. 5 and 6). Past data has suggested that PKI in the native I form was unlikely to be a simple globular protein since the apparent molecular weight by gel exclusion chromatography (22,000) is markedly higher than that determined from the amino acid sequence (7829 (6)). The high resistance of PKI to conditions of both low pH and high temperature has also suggested that it may have minimal structure. The amount of material that can now be obtained by bacterial expression has allowed this to be addressed. The combined results of FTIR and CD provide strong evidence that an unusually large proportion of the PKI protein is present in random coil ( Figs.   7 and 8). Also present are a-helix and p structures, however, and this is compatible with our past data on the presence of some structure in the most active peptides derived from PKI (30). Given these data it would appear likely that the I' form of PKI represents a more compact structure produced artifactually.
Bacterially produced PKI should provide a useful tool for studies of the catalytic subunit of the CAMP-dependent protein kinase. Natural PKI has been derivatized with fluorescein isothiocyanate and used as a probe to localize catalytic subunit in cultured cells (31,32). Recent studies suggest that the recombinant PKI can be useful for isolation of the catalytic subunit from mammalian cell extracts (33). The ability to readily alter the PKI coding sequence should also permit further analysis of PKI structure/function. While this area has already been substantially explored using synthetic peptides (7-9), most studies to date have concentrated on the amino-terminal region which is required for biological activity of PKI. However, it seems likely that additional residues beyond this region facilitate the high affinity and selectivity of the intact PKI (10). The ability to readily alter the coding sequence of PKI coupled with the bacterial expression system should permit the functional analysis of other regions of the protein. Finally, the availability of relatively large amounts of the bacterially produced PKI may further facilitate structural studies including x-ray crystallographic studies of the inhibitor-catalytic subunit complex. Recent studies have demonstrated the expression of regulatory (34) and the catalytic (35) subunit of the CAMP-dependent protein kinase in E. coli. The regulatory subunit bacterial expression system has been used to explore structure/function questions (36-38). The ability to manipulate PKI sequences in bacterial expression vectors provides additional approaches to understanding the functions of the CAMP-dependent protein kinase.
Acknowledgments-We thank Dr. Stanley Tabor for generously providing the T7 RNA polymerase expression vectors which were used in this study. We thank B. Maurer for aid in preparing this manuscript. Construetlon of PKI expresrlon vector. Reparation of PKI coding sequences for vector was prepared by modlfying the previously described eukaryotic expression vector 112).
Insemon In the T 7 7 expression vector are summanred In Figure 3. me bacterial expresslon whlch contained the complete rabbit skeletal muscle coding Sequence prepared synthetlcally by reverse translauon of the m l n o acld sequence. Wth a n added methlanme inltiauon altered so that a Nco I Slte containlng t h e translatlon Inltiauon codon was converted to an codon and an amino termlnal glydne codon The PKI sequence of the eukaryotic vector was Nde I slte. Thls placed the lnltiatlon codon In the proper relauonshlp to the ribosome blndlng slte of the expresslon vector and should permlt the synthesls of full-length rahblt muscle PKI wlth only the addition of an lnltlator methlonlne plus a glyclne resldue. The T77~PKI DNA was Inltlally used to transform the DH-5a straln of E. coli. After characterizauon of the cloned DNA by resmction endonuclease dlgestian. DNA from an appropriate clone was used to transform strain CP1-2 for expresslon of PKl.
Assay of PKI actlvlty. The activity oi punned PKI was assayed by the inhlbitlon of the catalyuc subunit of the cAMP-dependent protein klnase as descrtbed by Whitehouse and Walsh (161. During pudfication. fractions were assayed far the ablllty to Inhlblt the actlvlty of parually purified catalytic subunlt of the cAMP-dependent protem kinase ln a 50 pl reaction containing 0.5 unlts of parually purified catalytic subunlt. 25 mM r n s~c l (pH 7.41. 5 mM magnesium acetate. 5 mM dlthlothreltol. 20 pM Kemptlde and 0.1 mM IpPlATP 1200 cpmlpmoll. Reactions Were Incubated for 10 mln at 30% and 25 pI allquots spotted onto phosphocellulose ships and the filters washed 3 times with 75 mM phosphoric acld and radloaetivity detemlned by sclntlllation spectrometry as described by Roskoskr frozen ln an alcoholldry lce bath After thawing. lysozyme was added to a concentration 01 0.2 mg/ml and the cells Incubated for 45 mln on ~c e . After two freezeithaw ryrlea. cell debris was removed by centrifugahon at 30.000 ' p m for 20 min in a TI45 rotor IBeckmnn) The supernatant was removed and placed ln a bolllng water bath. The temperature of the cell extract was monltored. so that the emact was held at approximately 95°C for 5 mln. Denatured proteins were removed by centnfugatlon and the crude. heat-treated extract was adjusted to pH 5.0 wlth acetic acld After incubation for 30 mln at Toom tempcraturr.
an equal volume of 5 mM sodlum acetate. pH 5.0 and applied to a ZetaChrom 60 DEAI: Ion preclpitated proteins were removed by centrifugation and the supernatant was dllulrd Wth exchange dlsk. The DEAE dlsk was elutfd wth a slep gradlent of 0 1 M and 0.3 M sodium acetate. pH 5 0 . The protelns whlch were eluted with the 0.3 M sodium acetate were pod~tl. dlalyLed. lyophilized and fractionated by chromatography on a 2.5 x 75 cm column of Sephadex G50 In 0.01 M Ms (pH 7.5). The proteln content of the vaTlovs fractions was determined by colorimetric assay 1191 and t h e fractions were also examlned by SDS polyacrylamlde gel electrophoresls 120). FTIR mcasurcments. FTIR spectra were measured uslng a Mattson Alpha Centaun spectromcter at 4 cm r e~o l~t l o n .
To mlnlmlze the slgnal-to-nolse ratlo 1024 scans were c o~ added by triangularly apadlration The samples were In a Perkln-Elmer snlutlon cell wlth CaFz window and a 0.05 mm Teflon spacer. To mlnlmfze Interference by the rotational line structure bands from trace amounts of water vapor. the optical bench was mantamed under Constant poslUve dryN, pressure. Absorbance spectra were measured after a 25 mln purge and ratioed agalnst a slngle beam background specmm collected with no cdl. The spectrum of a buffer blank was subtracted from a sample spectrum to glve spectra free 01 contrrbutions from the buffer and cell. Kinase Inhibitor path length cell [Hellma) with a Jasca 40CS spectropolarimeter. The spectral bandwidth was CD measurements. CD spectra were measured a1 room temperature on a 0 01 Cm The spectra were measured for 2.5 h after a 20 min N2 purge. Samples were prepared by 1.0 nm. the sensltlvlty was 50 x 10 I m-tem. and Ume constants of 3 and 10 s were used. dllutlon of the protein samples prepared for FCIR measurements. DlluUonS were done by welghlng and the protein concentration for the spectra shown was 0.49 mgtml In 10 mM sodlum phosphate buffer. pH 7.0. Equkalent results were abtalned for PKI ~n bout phosphate and MOPS buffrr systems. The CD data were collected In 0.5 nm Increments over the range of 176.260 nm. The buffer data was subtracted from sample data and the resulung dlffcrence was smoothed uslng a Fourler smoothing routine "SMOOFT' 1211 wiUl a window wvldth of nine points. The secondary smcture conlcnt analysls software IFORTRAN 77 programs 'VARSELEC and "SEARCH") and the 22 protein data base were obtalnrd from W. C. Johnson [Deparlment of Blachemlstry and Biophysics. Oregon Slate Unwerslty. Cowallis. OR). The program 'VAHSELEC fitted the smoothed data over the range of 178-260 nm to linear comblnatlons of the Spectra from a set of 22 basis proteins takrn in groups of 19 [total number of groups was 15401. Following execuuon of 'VAIISELEC. the output file contalnlng the results Of each of the fits to all lhe I9-prateln groups was examined using "SEARCH with increasing strlngnt requlrements fRMS emor less than 0.18 and total secondary smcture between 90 and I lO%l unul five groups remalned. The spectra and secondary Smcture contcnl from those 19.proreln groups that satisfied the final search rcqulrements were averaged to produce lhe ntted spectra and the values for the secondary slmftum Content.