Substrate Requirements for Regulated Intramembrane Proteolysis of Bacillus subtilis Pro- K

JOURNAL OF BACTERIOLOGY, Feb. 2005, p. 961 971 Vol. 187, No. 3 0021-9193/05/$08.00 0 doi:10.1128/jb.187.3.961 971.2005 Copyright 2005, American Society for Microbiology. All Rights Reserved. Substrate Requirements for Regulated Intramembrane Proteolysis of Bacillus subtilis Pro- K Heather Prince, Ruanbao Zhou, and Lee Kroos* Department of Biochemistry and Molecular Biology, Michigan State University, East Lansing, Michigan Received 18 August 2004/Accepted 21 October 2004 During sporulation of Bacillus subtilis, pro- K is activated by regulated intramembrane proteolysis (RIP) in response to a signal from the forespore. RIP of pro- K removes its prosequence (amino acids 1 to 20), releasing K from the outer forespore membrane into the mother cell cytoplasm, in a reaction catalyzed by SpoIVFB, a metalloprotease in the S2P family of intramembrane-cleaving proteases. The requirements for pro- K to serve as a substrate for RIP were investigated by producing C-terminally truncated pro- K fused at different points to the green fluorescent protein (GFP) or hexahistidine in sporulating B. subtilis or in Escherichia coli engineered to coexpress SpoIVFB. Nearly half of pro- K (amino acids 1 to 117), including part of sigma factor region 2.4, was required for RIP of pro- K -GFP chimeras in sporulating B. subtilis. Likewise, pro- K -hexahistidine chimeras demonstrated that the N-terminal 117 amino acids of pro- K are sufficient for RIP, although the N-terminal 126 amino acids, which includes all of region 2.4, allowed much better accumulation of the chimeric protein in sporulating B. subtilis and more efficient processing by SpoIVFB in E. coli. In contrast to the requirements for RIP, a much smaller N-terminal segment (amino acids 1 to 27) was sufficient for membrane localization of a pro- K -GFP chimera. Addition or deletion of five amino acids near the N terminus allowed accurate processing of pro- K, ruling out a mechanism in which SpoIVFB measures the distance from the N terminus to the cleavage site. A charge reversal at position 13 (substituting glutamate for lysine) reduced accumulation of pro- K and prevented detectable RIP by SpoIVFB. These results elucidate substrate requirements for RIP of pro- K by SpoIVFB and may have implications for substrate recognition by other S2P family members. Bacillus subtilis is a gram-positive bacterium that undergoes sporulation when nutrients become limiting (reviewed in reference 57). Sporulation involves the formation of an asymmetrically positioned septum that divides the rod-shaped cell into a larger mother cell compartment and a smaller forespore (Fig. 1A). Subsequently, the septum migrates, engulfing the forespore and pinching it off as a free protoplast within the mother cell (Fig. 1B). Next, cell wall-like material is synthesized between the two membranes surrounding the forespore, and a coat composed of proteins made in the mother cell assembles on the surface of the forespore. The developmental process, called endosporulation, culminates with lysis of the mother cell to release a spore adapted for survival under harsh environmental conditions. Driving the endosporulation process are distinct but coordinated programs of gene regulation in the forespore and mother cell (reviewed in references 30, 32, 46, and 62). Much of the regulation is transcriptional, involving the synthesis and activation of sigma factors, which are subunits of RNA polymerase that enable it to recognize specific promoters. The first compartment-specific sigma factor to become active is F (Fig. 1A), which is released from an anti-sigma factor in the forespore shortly after asymmetric septum formation (11, 40, 41). Transcription by F RNA polymerase of spoiir in the forespore is believed to lead to secretion of the SpoIIR protein from the forespore (21, 27, 35), activating a protease, SpoIIGA, to cleave pro- E (23, 59). Both SpoIIGA and pro- E localize to * Corresponding author. Mailing address: Department of Biochemistry and Molecular Biology, Michigan State University, East Lansing, MI 48824. Phone: (517) 355-9726. Fax: (517) 353-9334. E-mail: kroos @msu.edu. the asymmetric septum (13, 20, 24). After cleavage, E accumulates only in the mother cell (15, 47) (Fig. 1A). There, E RNA polymerase transcribes many genes (12), including several required for engulfment of the forespore and activation of G (made in the forespore under F control) and one, sigk, that encodes pro- K, the inactive precursor of K (7, 37, 61). Completion of engulfment triggers activation of G in the forespore by an unknown mechanism (28). This initiates a second signal transduction pathway from the forespore to the mother cell (Fig. 1B and C). G RNA polymerase transcribes spoivb, and the SpoIVB serine protease is thought to be secreted from the forespore into the space between the two membranes surrounding the engulfed forespore (6, 19, 66). SpoIVB has been shown to cleave SpoIVFA (9), which forms a complex with SpoIVFB and BofA in the outer forespore membrane (OFM) (52). Cleavage of SpoIVFA has been proposed to dissolve the complex (9), freeing the SpoIVFB protease (51, 71) from its inhibitor, BofA (74). This would allow SpoIVFB to cleave pro- K associated with the OFM (72), releasing K to the mother cell cytoplasm (Fig. 1C). K and G RNA polymerases transcribe genes in the mother cell and forespore, respectively, whose products are needed for completion of endosporulation and mother cell lysis. The purpose of our investigation was to elucidate the requirements for pro- K to serve as a substrate for cleavage by SpoIVFB. This processing reaction is of special interest because it involves regulated intramembrane proteolysis (RIP) in response to a signal from the forespore. RIP is an important and widely conserved mechanism that governs signaling pathways in both prokaryotes and eukaryotes (3, 63). It involves cleavage of a protein within a membrane or near the membrane surface, releasing a polypeptide that typically acts as a 961

962 PRINCE ET AL. J. BACTERIOL. FIG. 1. Morphological changes during B. subtilis sporulation and intercompartmental signaling pathways that govern sigma factor activity. (A) Asymmetric septum formation divides the cell into two compartments, F becomes active in the forespore, and this leads to activation of E in the mother cell. (B) Later during sporulation, the process of engulfment pinches off the forespore within the mother cell, G becomes active in the forespore, and this leads to activation of K in the mother cell. (C) An expanded view of the signaling pathway that begins with G RNA polymerase-directed transcription of spoivb in the forespore and results in SpoIVFB-dependent RIP of pro- K. SpoIVB (not to be confused with SpoIVFB) is believed to cross the inner forespore membrane (IFM) and activate a complex of SpoIVFA, BofA, and SpoIVFB located in the outer forespore membrane (OFM). The topologies depicted for SpoIVFA, BofA, and SpoIVFB are based on analysis of lacz and phoa fusions in E. coli (16, 64). The depicted insertion of the N-terminal prosequence of pro- K in the OFM is speculative. The C terminus of each protein is labeled. RIP of pro- K by SpoIVFB releases K from the OFM, allowing K RNA polymerase to form in the mother cell. signal or a transcription factor. Proteases that carry out RIP have amino acids essential for catalysis located in transmembrane segments and are called intramembrane-cleaving proteases (I-Clips). I-Clips occur in large families that are typically conserved from archaea to humans and sometimes in bacteria. SpoIVFB is a member of the S2P family of I-Clips (51, 71), named after its founding member, the human site-2 protease (S2P) (48). S2P cleaves transcription factors that regulate cholesterol, fatty acid biosynthesis, and the response to unfolded proteins in the endoplasmic reticulum (48, 69). In zebra fish, S2P is required for cartilage development (55). In Escherichia coli, the S2P family member YaeL performs RIP of an antisigma factor, RseA, leading to activation of the E regulon in response to unfolded proteins in the periplasm (1, 25, 67). Likewise, a paralog of SpoIVFB in B. subtilis, YluC, performs RIP of anti-sigma factor RsiW, activating the W regulon in response to alkaline shock (56). The S2P family member Eep has been proposed to generate a peptide pheromone involved in mating of Enterococcus faecalis (2, 3). RIP of pro- K by SpoIVFB provides an opportunity to investigate how an enzyme in the S2P family recognizes its substrate, with implications for understanding substrate recognition in diverse signaling pathways and potentially for treating disease. Since nothing was known about the requirements for pro- K to serve as a substrate for RIP by SpoIVFB, we modeled our initial experiments after studies on pro- E. The N-terminal 28 amino acids of pro- E had been shown to be sufficient for processing of a pro- E - K fusion protein, and the N-terminal 52 to 55 amino acids of pro- E were sufficient for processing when fused to proteins unrelated to sigma factors (5, 24). These results suggested that the prosequence of pro- E (amino acids 1 to 27) contains sufficient information for recognition by the putative SpoIIGA protease. Therefore, we initially tested whether the N-terminal 27 amino acids of pro- K are sufficient for RIP when fused to green fluorescent protein (GFP). This pro- K -GFP fusion protein contained the prosequence plus seven additional amino acids of pro- K. We found not only that the prosequence is insufficient for RIP by SpoIVFB but that nearly half of pro- K is required for RIP of pro- K -GFP fusion proteins. Hence, the substrate requirements are different for SpoIVFB than for SpoIIGA, indicating that these proteases function differently, despite both being membrane proteins (45, 49) and both cleaving membrane-associated pro-sigma factors (20, 24, 72). We also found that addition or deletion of five amino acids near the N terminus of the prosequence did not alter RIP of pro- K, but a substitution of glutamate for lysine (reversing the charge of the amino acid side chain) at position 13 in the prosequence impaired accumulation of pro- K and prevented detectable RIP by SpoIVFB. We discuss possible implications of our results for substrate recognition by other S2P family members. MATERIALS AND METHODS Construction of plasmids. The plasmids constructed for this study are described in Tables 1 and 2, except psb1, which was made by inserting gfp (encoding GFP), generated by PCR using por267 (50) as a template, into BamHI- SalI-digested psgmu2 (14). psl1 (37), containing sigk (encoding pro- K ) generated by site-specific recombination that joins spoivcb to spoiiic (61), was used as a template for PCR to produce DNA fragments extending from 108 bp relative to the sigk transcriptional start site (34) to different points in the sigk coding region, which were inserted into EcoRI-BamHI-digested psb1 to create in-frame fusions to gfp. Likewise, PCR fragments containing the sigk promoter and portions of the sigk coding region were inserted into EcoRI-BamHI-digested php1 to create in-frame fusions to H6 (encoding hexahistidine). Mutations in sigk were created using mutagenic primers and the QuikChange sitedirected mutagenesis kit (Stratagene). In addition to the mutations described in Tables 1 and 2, php1 was used as a template for mutagenesis to change codon 13 of sigk, which codes for lysine, to instead code for glutamate or arginine, resulting in php40 and php41, respectively. All cloned PCR products and all genes subjected to mutagenesis were sequenced at the Michigan State University Genomics Technology Support Facility to confirm the presence of the desired sequences. Bacterial strains. B. subtilis strains were derived from BK410 (spoiiic94) (34). The spoiiic94 mutation deletes the 3 half (codons 114 to 241) of sigk (33), so BK410 fails to make pro- K (37). A BK410 derivative with a spoivf null mutation was constructed by transforming BK410 with chromosomal DNA from B. subtilis BMA2 (spoivf AB::spc), selecting on Luria-Bertani (LB) agar (54) containing spectinomycin (100 g/ml). The spoivf AB::spc mutation, which deletes the spoivf operon, was obtained by replacing the chloramphenicol resistance gene (cat) ofb. subtilis BSL51 (spoivf AB::cat) (38) with the spectinomycin resistance gene (spc) of pcm::sp as described previously (58). B. subtilis BK410 and its spoivf null mutant derivative were transformed with the plasmids described in Table 1, selecting on LB agar containing chloramphenicol (5 g/ml). Plasmids with different 3 ends of sigk fused to gfp or H6 were derived from psgmu2 (14) and were expected to integrate into the chromosome by homologous recombination (single crossover) with spoivcb, the 5 half (codons 1 to 113) of sigk (61). Transformants produced fusion proteins of the expected sizes when analyzed by Western blotting. Plasmids with an insertion or deletion mutation in the part of sigk encoding the prosequence were also expected to integrate into the chromosome by homologous recombination (single

VOL. 187, 2005 REQUIREMENTS FOR RIP OF PRO- K 963 TABLE 1. Plasmids transformed into B. subtilis Plasmid Description a psb2 Cm r ; sigk27-gfp; the sigk promoter and coding region from 108 bp to codon 27 was amplified by PCR, using psl1 as a template, and inserted into EcoRI-BamHI-digested psb1 pchc2 Cm r ; sigk47-gfp; same as psb2 except containing the sigk coding region to codon 47 prg2 Cm r ; sigk241-gfp; same as psb2 except containing the sigk coding region to codon 241 pdy1 Cm r ; sigk126-gfp; the sigk promoter and coding region from 108 bp to codon 126 was amplified by PCR, using prg2 as a template, and inserted into EcoRI-BamHI-digested psb1 pdy2 Cm r ; sigk166-gfp; same as pdy1 except containing the sigk coding region to codon 166 pdy3 Cm r ; sigk71-gfp; same as pdy1 except containing the sigk coding region to codon 71 pdy4 Cm r ; sigk94-gfp; same as pdy1 except containing the sigk coding region to codon 94 php1 Cm r ; sigk241-h6; annealed oligonucleotides coding for hexahistidine (H6) were inserted into BamHI-digested prg2, preserving a BamHI site upstream of H6 and introducing a stop codon downstream in-frame with gfp php5 Cm r ; sigk109-h6; the sigk promoter and coding region from 108 bp to codon 109 was amplified by PCR, using php1 as a template, and inserted into EcoRI-BamHI-digested php1 php6 Cm r ; sigk126-h6; the EcoRI-BamHI fragment from pdy1, containing the sigk promoter and coding region from 108 bp to codon 126, was subcloned into EcoRI-BamHI-digested php1 php7 Cm r ; sigk109-gfp; same as pdy1 except containing the sigk coding region to codon 109 php8 Cm r ; 2-6sigK126-H6; php6 was used as a template for mutagenesis to delete sigk codons 2 to 6 php9 Cm r ; 2-6sigK126-H6; php6 was used as a template for mutagenesis to insert a second copy of codons 2 to 6 after the first copy php11 Cm r ; 2-20sigK109-gfp; php7 was used as a template for mutagenesis to delete sigk codons 2 to 20 php18 Cm r ; sigk117-gfp; same as pdy1 except containing the sigk coding region to codon 117 php19 Cm r ; sigk117-h6; same as php5 except containing the sigk coding region to codon 117 php28 Cm r ; 2-6sigK241-H6; the SacI-HindIII fragment from php1, containing the 3 end of sigk fused to H6, was subcloned into SacI- HindIII-digested php8 php42 Cm r ; E113PsigK241-H6; php1 was used as a template for mutagenesis to change codon 113 of sigk, which codes for glutamate, to instead code for proline php44 Cm r ; 2-20sigK27-gfp; psb2 was used as a template for mutagenesis to delete sigk codons 2 to 20 php46 php47 a Cm r, chloramphenicol resistant. Cm r ; K13EsigK241-H6; the EcoRI-HindIII fragment of php40, containing K13EsigK241-H6, was subcloned into EcoRI-HindIII-digested pdg364 Cm r ; K13RsigK241-H6; the EcoRI-HindIII fragment of php41, containing K13RsigK241-H6, was subcloned into EcoRI-HindIII-digested pdg364 crossover) with spoivcb, and recombination could occur upstream or downstream of the mutation in the plasmid. Recombination upstream of the mutation would produce the mutant form of pro- K. Such transformants were identified by colony PCR with upstream primer 5 -CAATGTATGGGCGCTTGATGAAG-3 and downstream primer 5 -TACTAAAAAGACAAGCTCTTTAACAAC-3, which amplify a segment of the chromosome from slightly farther upstream of the sigk promoter region than is present in the plasmids to slightly downstream of the prosequence, producing a product with different mobility in 2% agarose gel electrophoresis than transformants with a wild-type prosequence, in which recombination must have occurred downstream of the mutation. For the plasmid with the E113P change in full-length sigk-h6 (php42), recombination (single crossover) with spoivcb was expected to occur upstream of the mutation in all transformants, since DNA downstream of the mutation corresponds to spoiiic, the 3 half (codons 114 to 241) of sigk, which is missing in BK410 and its spoivf null mutant derivative. Plasmids with the K13E or K13R change in full-length sigk-h6 (php46 and php47, respectively) were derived from pdg364 (26), which permits gene replacement of amye in the chromosome by homologous TABLE 2. Plasmids transformed into E. coli recombination (double crossover). Transformants that were amye mutant were identified by loss of amylase activity on 1% potato starch medium stained with Gram s iodine solution as described previously (18). B. subtilis BK410 and its spoivf null mutant derivative bearing integrated plasmids (php1 or php28) were transduced with phage SP ::gere-lacz as described previously (8), selecting on LB agar containing erythromycin (1 g/ml) and lincomycin (25 g/ml). E. coli strains were derived from BL21(DE3) (Novagen), which can be induced to synthesize T7 RNA polymerase. BL21(DE3) was transformed with the plasmids described in Table 2, selecting on LB agar containing kanamycin sulfate (50 g/ml), to create control strains that overproduce only wild-type or mutant pro- K upon induction of T7 RNA polymerase. To create strains that also overproduce H10-SpoIVFB-GFP, BL21(DE3) was cotransformed with pzr2 and each plasmid in Table 2, as described previously (74). BL21(DE3) derivatives containing pzr8 [overproduces pro- K (1 109)-H6] or pzr12 [overproduces pro- K (1 126)-H6] alone or in combination with pzr2 have been described previously (74). Plasmid Description a php22 Km r ; T7-sigK117-H6; the SacI-HindIII fragment from php19, containing the 3 end of sigk fused to H6, was subcloned into SacI-HindIII-digested pzr12 php30 Km r ; T7-K13EsigK126-H6; pzr12 was used as a template for mutagenesis to change codon 13 of sigk, which codes for lysine, to instead code for glutamate php32 Km r ; T7-E14KsigK126-H6; pzr12 was used as a template for mutagenesis to change codon 14 of sigk, which codes for glutamate, to instead code for lysine php33 Km r ; T7-E113AsigK126-H6; pzr12 was used as a template for mutagenesis to change codon 113 of sigk, which codes for glutamate, to instead code for alanine php34 Km r ; T7-E113PsigK126-H6; same as php33 except codon 113 of sigk was changed to code for proline php38 Km r ; T7-H117AsigK126-H6; pzr12 was used as a template for mutagenesis to change codon 117 of sigk, which codes for histidine, to instead code for alanine pzr98 Km r ; T7-2-6sigK126-H6; pzr12 was used as a template for mutagenesis to insert a second copy of sigk codons 2 to 6 after the first copy pzr100 Km r ; T7-2-6sigK126-H6; pzr12 was used as a template for mutagenesis to codons 2 to 6 a Km r, kanamycin resistant.

964 PRINCE ET AL. J. BACTERIOL. Cell growth and sporulation. E. coli and B. subtilis were typically grown on LB medium (54). Sporulation of B. subtilis was induced by growing cells in the absence of antibiotic and resuspension of cells in SM medium as described previously (18). The onset of sporulation is defined as the time of resuspension. Western blot analysis. For B. subtilis, samples (1 ml) were collected at the indicated times after the onset of sporulation, cells were collected by centrifugation (12,000 g), the supernatant was removed, and the cell pellet was stored at 80 C. For E. coli, equivalent amounts of cells from different cultures were collected from 0.5 to 1.0 ml of culture (depending on the optical density at 600 nm,) by centrifugation (12,000 g). Unless otherwise specified, whole-cell extracts were prepared as described previously (74). Proteins in extracts were separated by sodium dodecyl sulfate (SDS)-polyacrylamide gel electrophoresis and subjected to Western blot analysis as described previously (31), except that pro- K -GFP and K -GFP were separated on a lower-percentage (10%) polyacrylamide gel. Exposure times were 0.5 to 1 min, unless otherwise noted. Subcellular fractionation. The method described previously (72) was used except that the cell pellet was resuspended in 5% of the original volume of lysis buffer, cells were lysed by sonication, cell debris was removed immediately by low-speed centrifugation (12,000 g) for 1 min at 4 C, the supernatant was subjected to high-speed centrifugation (200,000 g) for1hat4 C, and the pellet (membrane fraction) was rinsed twice and then resuspended in 1/10 the lysate volume of lysis buffer without enzymes. The volume of the membrane fraction was then adjusted to equal the volume of the supernatant after high-speed centrifugation, which was the cytoplasmic fraction. Membrane and cytoplasmic fractions, as well as a sample of the supernatant after low-speed centrifugation, were subjected to Western blot analysis. Purification of processed proteins and determination of amino acid sequence. E. coli BL21(DE3) containing pzr2 and pzr98 or pzr100 was induced to produce proteins as described previously (74) in 25 ml of LB medium. Unprocessed pro- K (1-(2 6) 2-126)-H6 or pro- K (1-[ 2 6]-126)-H6, together with the corresponding processed protein, was purified from a cell extract using cobalt affinity chromatography (Clontech) as described by the manufacturer. The unprocessed and processed proteins were separated on SDS 14% Prosieve polyacrylamide gels (Cambrex Bio Science, Rockland, Maine) with Tris-Tricine electrode buffer (0.1 M Tris, 0.1 M Tricine, 0.1% SDS), electroblotted to Sequi-Blot polyvinylidene difluoride membranes (Bio-Rad), stained with Coomassie solution (0.1% Coomassie brilliant blue R-250, 1.0% acetic acid, 40% methanol), and destained with 50% methanol, and the faster-migrating, processed protein was sequenced by Edman degradation at the Michigan State University Macromolecular Structure Facility. FIG. 2. Processing of pro- K fusion proteins with C-terminal tags. The top part shows a map of pro- K, with amino acid numbers above the line indicating the boundaries of factor regions (36) listed below the line. The bottom part shows Western blot analysis of pro- K fusion proteins expressed from genes integrated by homologous recombination into the chromosome of B. subtilis BK410 or its processing-deficient spoivf null mutant derivative. Samples were collected at the indicated number of hours after induction of sporulation. Whole-cell extracts were subjected to Western blot analysis with antibodies against GFP or pentahistidine except that full-length pro- K -GFP and its processed form, K -GFP, were detected with antibodies against pro- K. Rightward arrows point to unprocessed fusion proteins. Leftward arrowheads point to processed fusion proteins. The solid leftward arrow points to proless- K (1-[ 2-20]-109)-GFP, used as a control to show that the processed form of pro- K (1-109)-GFP, if formed, would not comigrate with the faster-migrating species (dashed leftward arrow), which presumably results from degradation of pro- K (1-109)- GFP. Exposure times for the Western blots were 0.5 to 1 min, except for the pro- K (1-117)-H6 blot, which was a 60-min exposure. RESULTS RIP of pro- K -GFP fusion proteins. To test whether the prosequence of pro- K is sufficient for RIP, we fused DNA coding for the first 27 amino acids of pro- K in-frame with gfp. This sigk27-gfp fusion included DNA upstream of sigk to facilitate homologous recombination into the chromosome after transformation into B. subtilis BK410, which is missing the 3 half of the sigk gene and therefore fails to make pro- K (33, 37). As a control, the integrating plasmid bearing the sigk27- gfp fusion was transformed into a BK410 derivative with a spoivf null mutation, which prevents RIP of pro- K when present in otherwise wild-type cells (data not shown). Both strains accumulated pro- K (1 27)-GFP at 3 to 6 h after the initiation of sporulation, as detected by Western blot analysis of whole-cell extracts with anti-gfp antibodies, and there was no evidence of spoivf-dependent RIP of pro- K (1 27)-GFP (Fig. 2). Since the first 55 amino acids of pro- E are sufficient for processing of a pro- E -GFP fusion protein (24), we tested a comparable pro- K -GFP fusion protein. Based on an alignment of factor amino acid sequences, residue 55 of pro- E corresponds to residue 47 of pro- K (36). Therefore, we constructed sigk47-gfp and tested for spoivf-dependent RIP as described above. Like pro- K (1 27)-GFP, pro- K (1 47)-GFP accumulated during sporulation but was not processed (data not shown). Similarly, neither pro- K (1 71)-GFP, which includes not only region 1.2 but also region 2.1 of pro- K (Fig. 2), nor pro- K (1 94)-GFP, which further includes region 2.2, was processed (data not shown). We next considered the possibility that GFP was incompatible with RIP of pro- K. The gfp gene was fused in-frame at the 3 end of full-length sigk to create sigk241-gfp. This fusion gene, upon integration into the chromosome of the B. subtilis sigk mutant BK410, restored its ability to form heat-resistant spores but failed to confer this Spo phenotype on the processing-deficient spoivf mutant derivative of BK410. Consistent with these observations, pro- K -GFP was processed to a slightly smaller form beginning at h 4 of sporulation in BK410 but not in its spoivf mutant derivative (Fig. 2). We conclude that pro- K -GFP is processed to functional K -GFP. Convinced that GFP was compatible with RIP of pro- K,we fused gfp in-frame at the end of region 3 or region 2 of pro- K, creating sigk166-gfp and sigk126-gfp, respectively. Both pro- K (1 166)-GFP (data not shown) and pro- K (1 126)-GFP (Fig. 2) were processed beginning at h 4 of sporulation in a spoivfdependent fashion. To further localize the region of pro- K required for RIP,

VOL. 187, 2005 REQUIREMENTS FOR RIP OF PRO- K 965 FIG. 3. Processing of pro- K -H6 fusion proteins in E. coli. (A) Western blot analysis of pro- K (1 126)-H6 (lanes 1 to 3) and pro- K (1 117)-H6 (lanes 4 to 6) overproduced in E. coli bacteria that also overproduce H10-SpoIVFB-GFP. Samples were collected 1 h after induction of protein production from three independent isolates of each strain. Whole-cell extracts were prepared from equivalent amounts of cells based on culture optical density and subjected to Western blot analysis with antibodies against pentahistidine. Arrows point to unprocessed fusion proteins. Arrowheads point to processed fusion proteins. (B) Western blot analysis of whole-cell extracts (W) and supernatant fractions (S) after low-speed centrifugation (12,000 g) of pro- K -H6 fusion proteins overproduced in E. coli with or without H10- SpoIVFB-GFP. Samples were collected 1 h after induction of protein production. Fusion proteins were detected with antibodies against pentahistidine. Arrows and arrowheads have the same meaning as in (A). gfp was fused in-frame at the end of region 2.3. The resulting pro- K (1 109)-GFP did not appear to be processed in a spoivf-dependent fashion; however, a faster-migrating species was observed at all times tested during sporulation (Fig. 2), which could conceivably have masked our ability to detect the processed form of pro- K (1 109)-GFP. Therefore, we deleted DNA coding for amino acids 2 to 20 of pro- K (1 109)-GFP, creating 2-20sigK109-gfp. Figure 2 shows that proless- K (1- [ 2 20]-109)-GFP accumulated at h4ofsporulation and did not comigrate with the faster-migrating species observed in cells producing pro- K (1 109)-GFP. We conclude that pro- K (1 109)-GFP is not processed in a spoivf-dependent fashion. In contrast, when we fused gfp in-frame midway between the end of region 2.3 and 2.4 to create sigk117-gfp, the fusion protein was processed beginning at h 4 of sporulation in a spoivf-dependent fashion (Fig. 2). Taken together, the results shown in Fig. 2 demonstrate that nearly half of pro- K is required for spoivf-dependent RIP of pro- K -GFP fusion proteins. RIP of pro- K -histidine-tagged fusion proteins. Using a similar strategy, we constructed a series of fusions designed to produce C-terminally truncated pro- K proteins tagged with hexahistidine (H6), to see if this small tag would alter the results. Consistent with the GFP fusions, pro- K -H6 (data not shown) and pro- K (1 126)-H6 (Fig. 2) accumulated and were processed beginning at h 4 of sporulation in a spoivf-dependent manner. Likewise, pro- K (1 117)-H6 was processed, but it accumulated poorly, requiring a much longer exposure to be detected by Western blotting (Fig. 2). Pro- K (1 109)-H6 did not accumulate to a detectable level (data not shown). These results indicate that the region between amino acids 109 and 126 is important for accumulation of pro- K -H6 fusion proteins and support the conclusion that the N-terminal 117 amino acids of pro- K is sufficient for RIP. Processing of pro- K -H6 fusion proteins in E. coli. We reported recently that pro- K -H6 and pro- K (1 126)-H6 can be accurately and abundantly processed in E. coli engineered to overproduce one of these substrates and H10-SpoIVFB-GFP, a doubly tagged SpoIVFB chimera (74). Using this T7 RNA polymerase expression system in E. coli, we found that more pro- K (1 117)-H6 accumulated than in sporulating B. subtilis, but very little was processed in comparison with pro- K (1 126)-H6 (Fig. 3A). We investigated this difference by subjecting the whole-cell extracts to low-speed centrifugation (12,000 g) in order to separate insoluble protein aggregates, which form a pellet, from cytoplasmic and membrane-associated proteins, which remain in the supernatant. Figure 3B shows that very little, if any, pro- K (1 117)-H6 remained in the supernatant after low-speed centrifugation, whereas roughly half of the pro- K (1 126)-H6 remained in the supernatant. Figure 3B also shows that processing was dependent on coexpression of H10-SpoIVFB-GFP and that the processed forms were predominantly in the supernatant. We infer that the feeble processing observed when pro- K (1 117)-H6 was overproduced in E. coli resulted from inaccessibility of the substrate to the H10-SpoIVFB-GFP protease. Pro- K (1 109)-H6 accumulated to a barely detectable level in E. coli, and there was no evidence of processing (Fig. 3B). In E. coli, the region between amino acids 109 and 117 is important for accumulation of pro- K -H6 fusion proteins, and the region between amino acids 117 and 126 is important to prevent insoluble protein aggregates from forming, although the N terminal 117 amino acids of pro- K are sufficient for processing, as in sporulating B. subtilis. Effects of mutations that change amino acids in region 2.4 of pro- K. Having demonstrated that the region between amino acids 109 and 117 is necessary for spoivf-dependent RIP of pro- K -GFP fusion proteins and that this region is important for accumulation of pro- K -H6 fusion proteins both in sporulating B. subtilis and in growing E. coli, we sought to define critical amino acid residues and secondary structural features of this region. Figure 4A shows an alignment of this region with the corresponding region of B. subtilis pro- E, which is believed to be processed by SpoIIGA (23, 59), and Clostridium difficile K, which does not appear to be processed (17). The three sequences are identical except at the position corresponding to histidine 117 of pro- K. On the other hand, the entire region, including histidine 117, is conserved in pro- K orthologs of closely related Bacillus species (Baccilus cereus, Baccilus anthracis, and Baccilus halodurans). To test whether the side chain of histidine 117 plays a role in processing, we mutated the DNA coding for this amino acid to instead code for alanine, in the context of pro- K (1 126)-H6, and overproduced this protein in E. coli, with or without coexpression of H10-SpoIVFB-GFP. The mutant protein was processed in a spoivfb-dependent fashion, indistinguishable from that of wild-type pro- K (1 126)-H6 (Fig. 4B), demonstrating that histidine 117 is not critical for processing. The crystal structures of fragments of E. coli 70 (39) and Thermus aquaticus A (4) show that the C-terminal part of

966 PRINCE ET AL. J. BACTERIOL. FIG. 4. Effects of amino acid substitutions at positions 117 and 113 of pro- K. (A) Alignment of amino acids 109 to 117 of pro- K (61) with the corresponding regions of B. subtilis pro- E (60) and C. difficile K (17). (B) Western blot analysis of whole-cell extracts (W) and supernatant fractions (S) after low-speed centrifugation (12,000 g) of H117A mutant (lanes 1 to 4) and wild-type (lanes 5 and 6) pro- K (1 126)-H6 fusion proteins overproduced in E. coli with or without H10-SpoIVFB-GFP. Samples were collected 1 h after induction of protein production. Fusion proteins were detected with antibodies against pentahistidine. Arrows point to unprocessed fusion proteins. The arrowhead points to processed fusion proteins. (C) Western blot analysis of whole-cell extracts of E113A (lanes 1 and 2) and E113P (lanes 3 and 4) mutant pro- K (1 126)-H6 fusion proteins overproduced in E. coli with or without H10-SpoIVFB-GFP. Samples were collected 1 h after induction of protein production. Fusion proteins were detected with antibodies against pentahistidine. Arrows and arrowheads have the same meaning as in (B). (D) Western blot analysis of E113P mutant pro- K -H6 fusion protein expressed from a gene integrated by homologous recombination into the chromosome of B. subtilis BK410 or its processing-deficient spoivf null mutant derivative. Samples were collected at the indicated number of hours after induction of sporulation. Whole-cell extracts were subjected to Western blot analysis with antibodies against pentahistidine. The arrow and arrowhead point to unprocessed and processed fusion protein, respectively. region 2.3 and the N-terminal part of region 2.4 form an -helix. The sequence of pro- K in these regions is compatible with the formation of an -helix, according to several secondary structure prediction programs and a homology model of amino acids 21 to 126 based on the T. aquaticus structure (data not shown), so it was attractive to think that this putative -helix might be important for accumulation and processing of pro- K. Glutamate 113 is midway between amino acids 109 and 117, which defined a region important for accumulation and processing of pro- K fusion proteins (Fig. 2 and 3). As a control, we changed this amino acid to alanine in the context of pro- K (1 126)-H6, and there was little or no effect on accumulation or processing in E. coli (Fig. 4C). Changing glutamate 113 to proline, designed to break the putative -helix, impaired accumulation of the unprocessed form, yet a small amount of the processed form accumulated (Fig. 4C). The reduced accumulation of the unprocessed form is consistent with the idea that the region forms an -helix important for the stability of the fusion protein. The processed form of the mutant protein is presumably also less stable than the processed form of wildtype pro- K (1 126)-H6 (Fig. 4B) or its E113A mutant derivative (Fig. 4C). Therefore, little of the processed form of the E113P mutant derivative would be expected to accumulate even if the mutation had no effect on the protein s ability to serve as a substrate for the processing reaction. To further investigate the effects of the E113P change, it was engineered into full-length pro- K -H6 in B. subtilis. Surprisingly, the mutant protein accumulated and was processed normally (Fig. 4D). If this region of pro- K forms an -helix, either this secondary structure is not a critical feature or a proline at position 113 does not disrupt it sufficiently in the context of the full-length protein to impair accumulation or RIP in sporulating B. subtilis. Membrane localization of pro- K fusion proteins. The majority of pro- K in sporulating B. subtilis is membrane associated (72). Immunofluorescence microscopy suggested that pro- K was localized to the mother cell membrane and the OFM. This localization was independent of the SpoIVFB protease, which appears to specifically localize to the OFM midway during sporulation. One possible role of the region between amino acids 109 and 117 of pro- K might be to facilitate membrane association. To test this possibility, we examined membrane association of pro- K fusion proteins by subcellular fractionation. Figure 5 shows that pro- K -H6 is found almost exclusively in the membrane fraction of spoivf null mutant B. subtilis. This is consistent with results observed previously for pro- K (72). Figure 5 shows that pro- K (1 27)-GFP is also predominantly membrane associated. As a control, 2-20sigK27- gfp was constructed to express proless- K (1-[ 2 20]-27)-GFP, and this protein was found predominantly in the cytoplasmic fraction (Fig. 5). These results demonstrate that the prosequence of pro- K is necessary, and the first 27 amino acids of pro- K are sufficient, for membrane localization of GFP. Moreover, both pro- K (1 109)-GFP and pro- K (1 126)-GFP were predominantly membrane associated (Fig. 5), so the inability FIG. 5. Membrane localization of pro- K fusion proteins. Western blot analysis of pro- K fusion proteins expressed from genes integrated by homologous recombination into the chromosome of B. subtilis BK410 with a spoivf null mutation. Samples were collected 4 h after induction of sporulation. Supernatant fractions (S) after low-speed centrifugation (12,000 g) and cytoplasmic (C) and membrane (M) fractions after high-speed centrifugation (200,000 g) were prepared as described in Materials and Methods. Fusion proteins were detected with antibodies against pentahistidine or GFP.

VOL. 187, 2005 REQUIREMENTS FOR RIP OF PRO- K 967 FIG. 6. Effects of a five-amino-acid insertion or deletion in the prosequence of pro- K. Western blot analysis of pro- K (1-[2 6] 2-126)- H6 and pro- K (1-[ 2 6]-241)-H6 proteins expressed from genes integrated by homologous recombination into the chromosome of B. subtilis BK410 or its processing-deficient spoivf null mutant derivative is shown. Samples were collected at the indicated number of hours after induction of sporulation. Whole-cell extracts were subjected to Western blot analysis with antibodies against pentahistidine. The arrows and arrowheads point to unprocessed and processed fusion proteins, respectively. of sporulating B. subtilis to process pro- K (1 109)-GFP is not due to a general inability to associate with membranes. Effects of changes in the prosequence of pro- K. To determine whether a longer prosequence is compatible with RIP, we inserted an extra segment of DNA in sigk126-h6, coding for a second copy of amino acids 2 to 6 of pro- K, in tandem with the first copy. The protein, designated pro- K (1-(2 6) 2-126)- H6, appeared to be processed in B. subtilis beginning at h 4 of sporulation in a spoivf-dependent fashion (Fig. 6). Likewise, E. coli engineered to overproduce pro- K (1-(2 6) 2-126)-H6 produced a faster-migrating species only when H10-SpoIVFB- GFP was also overproduced (data not shown). This species comigrated on SDS-polyacrylamide gels with K (21 126)-H6 (data not shown), the processed form of pro- K (1 126)-H6 (74). The processed form of pro- K (1-(2 6) 2-126)-H6 was abundant enough to permit purification and N-terminal amino acid sequencing by Edman degradation. The first five amino acids, YVKNN, matched the N-terminal amino acid sequence of K produced in sporulating B. subtilis (29) and that of K (21 126)- H6 produced in E. coli (74). Therefore, processing occurred at the normal site even when the prosequence was lengthened by five amino acids near the N-terminal end. This rules out a mechanism in which the SpoIVFB protease measures the distance from the N terminus to the cleavage site. To determine whether amino acids 2 to 6 of pro- K are important for processing, we deleted the DNA coding for these residues in sigk126-h6. The protein, designated pro- K (1-[ 2 6]-126)-H6, appeared to be processed in B. subtilis in a spoivfdependent fashion, and the processed form comigrated with K (21 126)-H6 (data not shown). E. coli engineered to overproduce pro- K (1-[ 2 6]-126)-H6 produced a faster-migrating species only when H10-SpoIVFB-GFP was also overproduced (data not shown), and N-terminal amino acid sequencing revealed that processing had occurred at the normal site. This further demonstrates that the SpoIVFB protease does not measure the distance from the N terminus to the cleavage site and shows that amino acids 2 to 6 of the prosequence are not required for processing. Deletion of the first six amino acids of pro- K had a dramatic effect on activity of the protein in vitro (22). The isolated protein bound DNA with higher affinity than pro- K or K, and in combination with E. coli core RNA polymerase, it stimulated transcription 30-fold more efficiently than K under high-salt conditions (250 mm KCl). To examine the effects of deleting amino acids 2 to 6 of pro- K, both in B. subtilis capable of processing and in a spoivf null mutant background incapable of processing (in order to see if the unprocessed protein has sigma factor activity in vivo), we constructed 2-6sigK241- H6. We found that the resulting protein, designated pro- K (1-[ 2 6]-241)-H6, appeared to be processed in B. subtilis beginning at h 4 of sporulation in a spoivf-dependent fashion (Fig. 6). Consistent with the idea that K -H6 was being produced, the ability of the BK410 mutant to form phase-bright spores was restored, based on microscopic examination. Interestingly, phase-bright spore formation also appeared to be partially restored in the BK410 derivative bearing the spoivf null mutation and the 2-6sigK241-H6 allele. Since processing was not detectable (Fig. 6), this suggested that pro- K (1-[ 2 6]-241)-H6 might be active without processing. Table 3 shows that the BK410 spoivf null mutant bearing the 2-6sigK241- H6 allele produced 580-fold more heat-resistant spores than with the sigk241-h6 allele, although the number of spores was 360-fold less than for either allele in the processing-competent BK410 background without the spoivf null mutation. Also, the BK410 spoivf null mutant bearing the 2-6sigK241-H6 allele expressed a K -dependent gere-lacz fusion (8) during sporulation well above the level observed in the corresponding strain with the sigk241-h6 allele (Fig. 7A) but far below the level seen for both alleles in the processing-competent BK410 strain (Fig. 7B). Taken together, these results suggest that pro- K (1-[ 2 6]-241)-H6 has slight K activity prior to RIP and full activity after RIP. Amino acids 7 to 20 of pro- K, ALGFVVKELVFLVS, are predominantly residues with hydrophobic side chains. Two exceptions are lysine 13 and glutamate 14. To test whether the particular charge on the side chains of these amino acids is required for processing, we made mutations that would produce proteins with charge reversals. Pro- K (1 126)-H6 with glutamate 14 changed to lysine (E14K) appeared to be processed in E. coli when H10-SpoIVFB-GFP was produced (Fig. 8A). A switch from a negatively charged side chain to a positively charged side chain at position 14 seems inconsequential. In contrast, pro- K (1 126)-H6 with a K13E charge reversal did not appear to be processed in E. coli, although a small amount appeared to be degraded to a species that comigrated with K (21 126)-H6 even when H10-SpoIVFB-GFP was not present (data not shown). Therefore, we investigated the effect of the K13E charge reversal in the context of full-length pro- K -H6 in sporulating B. subtilis. Interestingly, this mutant protein TABLE 3. Heat-resistant spore formation of B. subtilis producing wild-type or mutant pro- K fusion proteins Relevant genotype No. of spores/ml a sigk241-h6... 1.3 10 8 sigk241-h6 spoivf... 6.2 10 2 2-6sigK241-H6... 1.3 10 8 2-6sigK241-H6 spoivf... 3.6 10 5 a The number of heat-resistant spores/ml at 24 h after the onset of sporulation was determined as described previously (18), except that cells were plated in LB soft agar. All strains were derived from BK410 or its spoivf null mutant derivative. BK410 is isogenic with wild-type B. subtilis PY79 (70), which made 1.3 10 8 spores/ml under these conditions.

968 PRINCE ET AL. J. BACTERIOL. FIG. 7. Expression of a K -dependent reporter during sporulation. (A) -Galactosidase activity from gere-lacz during sporulation of B. subtilis BK410 spoivf null mutants bearing the sigk241-h6 allele coding for pro- K -H6 (E) orthe 2-6sigK241-H6 allele coding for pro- K (1-[ 2 6]-241)-H6 ( ), determined as described previously (73) for two or three SP ::gere-lacz transductants of each strain. Points show the averages, and error bars show one standard deviation of the data. (B) Same as (A) except that pro- K processing-competent BK410 without the spoivf null mutation was used. accumulated poorly and RIP was not detectable (Fig. 8B, lanes 2 to 7). Subcellular fractionation revealed that the K13E mutant protein was membrane associated (data not shown). A conservative K13R change in full-length pro- K -H6 did not impair spoivf-dependent RIP in sporulating B. subtilis (Fig. 8B, lane 1; also data not shown). We conclude that the K13E charge reversal does not prevent pro- K from associating with membranes but has a profound effect on stability of the mutant protein in sporulating B. subtilis and appears to prevent processing in E. coli or B. subtilis. We infer that a negatively charged glutamate side chain at position 13 of pro- K perturbs its structure, making it more susceptible to degradation by other proteases and apparently incompatible with RIP by SpoIVFB. DISCUSSION RIP is an important mechanism that controls signaling pathways in both prokaryotes and eukaryotes (3, 63), but little is known about the substrate requirements for I-Clips in the S2P family. We have investigated the requirements for pro- K to serve as a substrate for RIP by SpoIVFB, a member of the S2P family of I-Clips (51, 71). We discovered that nearly half of pro- K is required for RIP of pro- K -GFP fusion proteins, including at least part of region 2.4, which is far in the primary sequence from the processing site between amino acids 20 and 21, although three-dimensional modeling based on the structure of T. aquaticus A domain 2 (4) suggests that these two parts of pro- K may be as little as 20 Å apart (data not shown). We also found that region 2.4 is important for accumulation and RIP of pro- K -H6-tagged proteins. Sigma factor regions 3 and 4 were not necessary for accumulation or RIP of pro- K - H6-tagged proteins in sporulating B. subtilis or growing E. coli. Our results elucidate substrate requirements for RIP of pro- K by SpoIVFB, with possible implications for substrate recognition by other S2P family members. Why might region 2.4 be important for stability and RIP of pro- K? It is likely to be part of a domain composed of four -helices, based on the crystal structures of an E. coli 70 fragment (39) and the T. aquaticus A domain 2 (4). The C- terminal part of region 2.3 and the N-terminal part of region 2.4 together form an -helix in these structures, as well as in RNA polymerase holoenzyme structures (42, 65). Loss of all or half of region 2.4 due to truncation at amino acid 109 or 117, respectively, might destabilize the domain structure, making the protein more susceptible to degradation. Fusing GFP onto the C termini at these positions resulted in fusion proteins that accumulated, and pro- K (1 117)-GFP was processed, but pro- K (1 109)-GFP was not processed (Fig. 2). This could mean that a side chain of one or more amino acids between residues 109 and 117 of pro- K is crucial for RIP, perhaps to interact with the SpoIVFB protease or another part of pro- K, such as the prosequence. We can rule this out for amino acids 113 and 117, since an alanine substitution at either position did not prevent RIP. Further mutational analysis will be needed to address whether other amino acid side chains in this region are critical. An alternative model is that the ability to form an -helix, rather than the amino acid sequence, is the important feature of the region between amino acids 109 and 117. The N-terminal amino acid sequence of GFP is not favorable for -helix formation, so pro- K (1 109)-GFP might not form the putative required -helix. A proline substitution at position 113 in the middle of the region, designed to break the putative -helix, reduced accumulation of pro- K (1 126)-H6 but did not prevent RIP (Fig. 4C). The reduced accumulation supported the idea that the region is -helical, but when tested in FIG. 8. Effects of charge reversals in the prosequence of pro- K. (A) Western blot analysis of whole-cell extracts of E14K mutant pro- K (1 126)-H6 overproduced in E. coli with or without H10-SpoIVFB- GFP. Samples were collected 1 h after induction of protein production. Fusion proteins were detected with antibodies against pentahistidine. The arrow and arrowhead point to unprocessed and processed fusion protein, respectively. The two lanes are from the same gel, but intervening lanes were removed from the image. (B) Western blot analysis of K13R (lane 1) and K13E (lanes 2 to 7) mutant pro- K -H6 fusion proteins expressed from genes integrated by homologous recombination at the amye locus in the chromosome of B. subtilis BK410 (lanes 1 to 4) or its processing-deficient spoivf null mutant derivative (lanes 5 to 7). Samples were collected at the indicated number of hours after induction of sporulation. Whole-cell extracts were subjected to Western blot analysis with antibodies against pro- K. The arrows and arrowhead point to unprocessed and processed fusion proteins, respectively. All lanes are from the same gel, but intervening lanes of the gel were removed between lanes 1 and 2 when the image was created.