Zinc Inhibition of Protein trans-Splicing and Identification of Regions Essential for Splicing and Association of a Split Intein*

Two important aspects of protein splicing were investigated by employing the trans-splicing intein from the dnaE gene of Synechocystis sp. PCC6803. First, we demonstrated that both protein splicing and cleavage at the N-terminal splice junction were inhibited in the presence of zinc ion. The trans-splicing reaction was partially blocked at a concentration of 1–10 μm Zn2+ and completely inhibited at 100 μm Zn2+; the inhibition by zinc was reversed in the presence of ethylenediaminetetraacetic acid. We propose that inactivation of Cys160 at the C-terminal splice junction by the chelation of zinc affects both the N-S acyl rearrangement and the transesterification steps in the splicing pathway. Furthermore,in vivo and in vitro assays were established for the determination of intein residues and regions required for splicing or association between the N- and C-terminal intein halves. N-terminal truncation of the intein C-terminal segment inhibited both splicing and association activities, suggesting this region is crucial for the formation of an interface between the two intein halves. The replacement of conserved residues in blocks B and F with alanine abolished splicing but allowed for association. This is the first evidence showing that the conserved residues in block F are required for protein splicing.

Protein splicing is a post-translational processing event, which involves the self-catalyzed excision of an internal protein segment, or intein, from a protein precursor with the concomitant joining of the flanking polypeptide sequences, the exteins (1,2). Sequence alignment reveals that an intein can be divided into three segments. The N-terminal region possesses ϳ120 -150 amino acid residues including highly conserved blocks A and B, whereas the C-terminal region is composed of ϳ35-50 residues containing conserved blocks F and G. Between the two terminal regions is an optional endonuclease domain, which has been found in a majority of inteins (3,4). In the case of protein trans-splicing, however, a functional intein is reconstituted from two non-covalently linked N-and C-terminal segments that are separately translated (5,6). The crystal structures of inteins from a vacuolar ATPase subunit of Saccharomyces cerevisiae (PI-SceI or Sce VMA intein), 1 Myco-bacterium xenopi GyrA (Mxe GyrA intein), and Pyrococcus furiosus ribonucleotide reductase (PI-PfuI) revealed that the Nand C-terminal regions form a horseshoe-shaped Hint (hedgehog and intein) domain (7)(8)(9). The intein structure contains an unusual ␤-fold with the splice junctions at the ends of two adjacent antiparallel ␤-strands, forming a catalytic pocket. The catalytic residues implicated in the splicing mechanism have been found in the conserved blocks, A and G, present at the two splice junctions (10 -16). Presumably, these residues directly participate in protein splicing by three concerted nucleophilic replacements (1,17) (Fig. 1). Other residues and regions that may be involved in assisting these catalytic reactions have yet to be identified, although a highly conserved histidine residue in block B has been found to be necessary for protein splicing (18). Furthermore, protein splicing requires the precise alignment of the two splice junctions to form the active site. Examination of the interaction between the N-and C-terminal intein halves would offer important insight into the tertiary folding that brings all the reacting groups in close proximity.
Recently, the crystal structure of a precursor protein containing the Sce VMA intein with the flanking native extein residues was solved (19). Remarkably, a zinc atom was found at the catalytic center of this splicing-deficient precursor protein.
The residues contributing to the zinc coordination include a cysteine following the C-terminal splice site, which presumably receives the N-extein domain by transesterification, a critical step in the complex splicing pathway. Thus, it is of particular interest to investigate the possibility that zinc may play a modulatory role during the process of protein splicing. However, the rapid processing of intein precursors presents a major obstacle to the conduction of such an investigation (10 -12).
In this report, an important advance in understanding the complexities of protein splicing is made possible by the manipulation of the trans-splicing intein involved in the maturation of the Synechocystis sp. DnaE protein (5,6). The naturally occurring split intein, consisting of an N-terminal segment of 123 residues and a C-terminal segment of 36 residues, can mediate efficient splicing in trans in foreign protein contexts (20,21). Similar to the Sce VMA intein, the sulfhydryl group of the cysteine residues presumably function as nucleophiles at both the N-and C-terminal splice junctions of the DnaE intein ( Fig. 1). In this study, we took advantage of the unique property of the trans-splicing Ssp DnaE intein to explore the effect of zinc on splicing activity. This trans-splicing system also allowed us to investigate the intein elements responsible for the interaction between the N-and C-terminal halves of a protein splicing element, independent of its catalytic activity. We demonstrated the correlation between the splicing or association functions of the DnaE intein and the reconstitution of Esche- ‡ These authors made equal contributions to this work. § To whom correspondence should be addressed: New England Biolabs, Inc., 32 Tozer Rd., Beverly, MA 01915. Tel.: 978-927-5054; Fax: 978-921-1350; E-mail: xum@neb.com. 1 The abbreviations used are: Sce VMA intein, the intein from the 69-kDa vacuolar ATPase subunit of S. cerevisiae; ALS, E. coli acetolactate synthase isoform II; CBD, chitin binding domain; DTT, 1,4-dithiothreitol; GST, glutathione S-transferase; I C , the 36 C-terminal amino acids of the Ssp DnaE intein; I N , the 123 N-terminal amino acids of the Ssp DnaE intein; IPTG, isopropyl-␤-D-thiogalactopyranoside; MBP, maltose-binding protein; Mxe GyrA intein, the intein from M. xenopi richia coli acetolactate synthase isoform II (ALS). Co-expression of the two DnaE intein fragments fused to two halves of ALS rescued E. coli ER2744, which lacks an active ALS, from growth inhibition by valine (22)(23)(24). The effect of amino acid substitutions or deletions on the splicing and association activities of the two intein halves were further characterized by an in vitro assay system.

EXPERIMENTAL PROCEDURES
Cells and Materials-E. coli ER2744 (fhuA2 glnV44 el4-rfbD1 relA1 endA1 spoT1 thi-1 ⌬(mcrC-mrr)114::IS10 lacZ::T7 gene1) contains a mutated ALS gene. 2 Plasmid Construction-pMEB10 is a derivative of pMEB4 (20) and expresses a fusion protein of E. coli maltose-binding protein (MBP or M) fused at its C terminus to the N-terminal 123-residue segment (I N ) of the Ssp DnaE intein. pBEL11 expresses a three-part fusion protein containing a N-terminal chitin binding domain (CBD) from Bacillus circulans followed by the C-terminal 36-residue fragment (I C ) of the Ssp DnaE intein and T4 DNA ligase (20). Two pBEL11 variants were created by substitution of the last intein residue, Asn 159 , or the first C-extein residue, Cys 160 , with an alanine residue by linker replacement using the unique NruI and AgeI sites. The coding sequence of E. coli ALS was split into two segments and fused in-frame to the N-and C-terminal halves of the Ssp DnaE intein (I N and I C ) as described by Sun et al. (23). All intein fusion constructs contain five native extein residues flanking each splice junction to enhance splicing efficiency (20,23). pEN10 was created by the in-frame fusion of the DNA fragment encoding the N-terminal 327 amino acid residues of ALS, or ALS(N) carrying the A26V mutation (24), to the N terminus of I N generating the ALS(N)-I N fusion gene. pKEC1 was constructed by fusing the DNA fragment encoding the C-terminal 221 amino acid residues of ALS, ALS(C), in frame to the C terminus of I C , creating the I C -ALS(C) fusion gene. Mutations in block B were introduced by site-directed mutagenesis (QuickChange site-directed mutagenesis kit, Stratagene, La Jolla, CA) using the following primers: T69A, 5Ј-CAGTAATCCGAGCTGC-CTCTGACCAC-3Ј and 5Ј-CGGTGGTCAGAGGCAGCTCGGATTACTG-3Ј; H72A, 5Ј-GAGCTACCTCTGACGCCCGCTTTTTAACCACCG-3Ј and 5Ј-CGGTGGTTAAAAAGCGGGCGTCAGAGGTAGCTC-3Ј. Sequential deletions from the C terminus of I N were created by polymerase chain reaction amplification using the T7 universal primer (New England Biolabs) and the following oligonucleotides: 3-amino acid deletion, I N ⌬3, 5Ј-GGGCCCTGCAGTCACCCAGCGTCAAGTAATGG-3Ј; I N ⌬9, 5Ј-GG-GCCCTGCAGTCAAAGACGATGGTTGTCAAGAGC-3Ј; I N ⌬16, 5Ј-GG-GCCCTGCAGTCAAAGAGCTTCTTCAGTTTGC-3Ј; I N ⌬23, 5Ј-CCCTG-CAGTCAAATATTTTCTAAAGTCAACAAG-3Ј; I N ⌬29, 5Ј-GGGCCCTG-CAGTCACAAGTCCAGTTGCCTAGC-3Ј; and I N ⌬45, 5Ј-CCCTGCAGT-CAATCGGTGGTTAAAAAGCGGTG-3Ј. The resulting polymerase chain reaction fragments were cloned in frame to the C terminus of ALS(N), replacing the wild-type I N sequence using the XbaI and PstI sites in pEN10. Deletion constructs spanning residues 125-130 from the N-terminal region of I C , following the translation initiation codon for Met 124 , were created by linker replacement using the XbaI and BglII sites in pKEC1. The mutagenesis linkers were formed by annealing appropriate complementary oligonucleotides. pKEC3 was created by introducing an NdeI site overlapping the translation initiation codon into pKEC1 using the XbaI and BglII sites. The linker was formed by annealing the oligonucleotide, 5Ј-CTAGAAATAATTTTGTTTAACTT-TAAGAAGGAGATATACATATGGTTAAAGTTATCGGTCGTA-3Ј, and its complement. The following oligonucleotides and their appropriate complements were used to introduce deletions spanning residues 125-136 by linker insertion into the NdeI and NheI sites in pKEC3: I C ⌬8, 5Ј-TATGCTGGGCGTGCAGCGCATCTTTGATATCGGTCTGCCGCAG-GACCATAACTTTCTG-3Ј; I C ⌬10, 5Ј-TATGGTTCAGCGCATCTTTGA-TATCGGTCTGCCGCAGGACCATAACTTTCTG-3Ј; I C ⌬12, 5Ј-TATGC-GCATCTTTGATATCGGTCTGCCGCAGGACCATAACTTTCTG-3Ј. Mutagenesis of the conserved residues in block F was carried out by linker substitution using the unique BglII and NheI sites in pKEC3. The ALS N-terminal coding sequence, in pEN10 or its variants, was replaced by the DNA sequence encoding MBP from pMEB10 using the NdeI and XhoI sites. The ALS C-terminal coding sequence was substituted with the DNA sequence encoding glutathione S-transferase (GST) (25), using the AgeI and PstI sites in pKEC3 to generate pEG3, which contained the wild-type I C or its variants.
Effect of Zinc Ion on Protein Splicing-MBP-I N fusion protein was expressed from pMEB10 and purified on an amylose column (New England Biolabs). The amylose column was washed with 100 ml of Buffer A (20 mM Tris, pH 7.0, 500 mM NaCl), followed by 100 ml of Buffer A containing 10 mM EDTA and then 300 ml of Buffer A. Purified MBP-I N was eluted in Buffer A containing 10 mM maltose. The CBD-I C -T4 DNA ligase (BI C L) fusion proteins and its mutant variants were purified by chitin beads in Buffer A, as described previously (20). Purified MBP-I N was added to aliquots of 0.43 M BI C L bound to chitin beads at ϳ0.53 M final concentration and incubated at 4°C for 24 h in various concentrations of zinc acetate (pH 6.5), or 1 mM calcium chloride (pH 6.8) or 1 mM magnesium chloride (pH 6.8). The samples were boiled in 3ϫ SDS sample buffer (New England Biolabs) and electrophoresed on a 12% Tris-glycine gel (Invitrogen, Carlsbad, CA) followed by staining with Coomassie Brilliant Blue. BI C L (N159A) at 1.7 M was incubated with MBP-I N at 2.0 M, in the absence or presence of 1 mM Zn 2ϩ at 4°C for 24 h. Cleavage mediated by MBP-I N and BI C L (C160A) was conducted in 8 M protein concentration with or without 5 mM 1,4dithiothreitol (DTT) at 4°C for 24 h. Structure Model-The M. xenopi GyrA intein crystal structure data was obtained from NCBI Structures data base (9) and used as a model for the Ssp DnaE intein. Fig. 3B was produced using Swiss-PdbViewer version 3.7b1. The structure corresponding to the region between positions Gly 132 and Tyr 162 of the Mxe GyrA intein is not shown.
Detection of Spliced Product by Western Blot Analysis-A single bacterial colony was inoculated in LB medium supplemented with 100 g/ml ampicillin and 50 g/ml kanamycin for 4 h at 37°C. Protein expression was induced with 0.3 mM IPTG for 16 h at 30°C. Whole cell lysates were resolved by SDS-PAGE. Western blots with the antibodies against the N-or C-terminal fragments of ALS were performed as described previously (23).
In Vitro Assays for Splicing and Association Activities-The chimeric proteins, MBP-I N and I C -GST, were mixed in vitro to assess the effect of mutations on splicing and the interaction of the two intein halves. The MBP-I N proteins were expressed and purified on amylose. The I C -GST fusion protein, expressed from pEG3 or its mutant derivative, was purified from the cell lysate by glutathione-Sepharose 4B resin (Amersham Pharmacia Biotech). The purified MBP-I N protein was added to I C -GST-bound resin. After a 45-min incubation at 4°C, the column was washed with at least 20 column volumes of Buffer A. Samples of glutathione-Sepharose resin were taken when MBP-I N was added to the resin (t 0 ), after incubation of 45 min, and after the resin was thoroughly washed. The samples were incubated at 4°C for 24 h before being analyzed by Coomassie Blue-stained SDS-PAGE.

Zinc Inhibits Protein trans-Splicing-The effect of zinc on
trans-splicing of the Ssp DnaE intein was examined by employing the in vitro trans-splicing system described previously (20) (Fig. 1). This system allowed the splicing reaction to occur between two bacterially expressed proteins, MBP-I N (MI N , 57 kDa) and CBD-I C -T4 DNA ligase (BI C L, 69 kDa), to yield the spliced product MBP-T4 DNA ligase (ML, 100 kDa). Analysis of the reaction by SDS-PAGE revealed that the spliced product, ML, was not observed when zinc ion is present at 1 mM concentration ( Fig. 2A), suggesting protein trans-splicing is inhibited. This inhibitory effect by zinc ion could be reversed by including 5 mM EDTA (metal chelating agent). The presence of either 1 mM Ca 2ϩ or 1 mM Mg 2ϩ appeared to have no significant effect on trans-splicing. These data indicated that the inhibitory effect of zinc ion on trans-splicing is not simply due to the presence of a divalent cation, but is specific to zinc. This inhibitory effect was observed in the presence of either zinc acetate or zinc chloride (data not shown). The effect of zinc ion on cleavage at the N-terminal splice junction, the initial step in the splicing pathway, was examined in the presence of 5 mM DTT. DTT-induced N-terminal cleavage is due to nucleophilic attack by thiol at thioester bond formed at Cys 1 (10,15). In the absence of Zn 2ϩ , both the spliced product ML and the N-terminal cleavage product (MBP or M, 42 kDa) were generated ( Fig. 2A, lane 8). In the presence of zinc, however, DTT induced N-terminal cleavage was abolished since no N-extein, M, was observed (lane 9).
To further investigate the inhibitory effect of zinc on transsplicing, MBP-I N and CBD-I C -T4 DNA ligase were incubated at 4°C for 24 h in various concentrations of zinc (Fig. 2B). The Coomassie Blue-stained SDS-PAGE showed that the effect of zinc on trans-splicing is concentration-dependent. The presence of 1 or 10 M Zn 2ϩ , at 2-20-fold in excess of protein concentration, resulted in partial inhibition of trans-splicing ( lanes 5 and 6). The amount of the spliced product, ML, decreased by ϳ85% in the presence of 10 M Zn 2ϩ , when com-pared with the yield in the absence of zinc (lane 4). However, in the presence of 100 M zinc ion, no ML was observed (lane 7), indicating that trans-splicing was inhibited.
We reasoned, based on the Sce VMA intein structure, that the inhibition of splicing and N-terminal cleavage is caused by the binding of the zinc ion to Cys 160 . This chelation will inhibit transesterification, thereby affecting the equilibrium between FIG. 1. Proposed mechanism for protein trans-splicing and its inhibition by zinc involving the Ssp DnaE intein. trans-Splicing is initiated by the association of the I N (123 amino acids) and I C (36 amino acids) of the Ssp DnaE intein. The folding of two associated intein halves aligns the two splice junctions in close proximity. The splicing reaction presumably occurs via the same splicing pathway as cis-splicing, proposed previously (10,14). Cys 160 participates in the formation of the branched intermediate by transesterification through a nucleophilic attack of the thioester bond formed between the last N-extein residue, and Cys 1 of the intein. Both protein splicing and cleavage at the Nterminal splice junction may be blocked by the chelation of a zinc atom involving the sulfur atom of Cys 160 at the C-terminal splice junction. amide and thioester, or the acyl shift at Cys 1 . We examined the effect of zinc in a splicing-deficient mutant carrying an N159A mutation. As shown in Fig. 2C, in the presence of 1 mM zinc no N-terminal cleavage product (M) was detected, whereas cleavage was observed in the absence of zinc. In addition, examination of N-terminal cleavage using a C160A mutant showed that no cleavage product was observed with or without DTT, implying that the inactivation of Cys 160 could affect acyl rearrangement, the first step of protein splicing (Fig. 2D).

Analysis of Intein Mutants by an in Vivo Functional Screen-
The sequence alignment of the Ssp DnaE intein with the Mxe GyrA intein and the Sce VMA intein is shown in Fig. 3A, along with the secondary structure assignments (3,4,8,9,19). Also presented in Fig. 3B is a model structure of the DnaE intein based on the x-ray structure of the Mxe GyrA intein (9). Two long and curved antiparallel ␤-strands (␤5 and ␤10) form part of two three-stranded mixed ␤-sheets (sheet 1: ␤4, ␤5, ␤10; sheet 2: ␤10, ␤5, ␤6). The N-terminal region of I C appears to align with the sequence elements participating in the formation of the ␤10 strand of the Mxe GyrA intein. A deletion in I C could disrupt formation of the three-stranded ␤-sheets and thus interaction between the two intein halves. The sequence spanning the C-terminal 24 residues of the I N segment appears to be a flexible linker region, which does not appear to be conserved among inteins (Fig. 3A). However, the sequence upstream of this region aligns with the amino acid residues of the GyrA intein that participate in the formation of the ␤9 strand, which appears to interact with the intein C-terminal segment (9). We hypothesized that a disruption of any of the ␤-sheets involved in formation of the ␤-core would affect splicing or association between the two intein fragments. In order to determine the regions required for the association and transsplicing functions of the DnaE intein, sequential deletions were introduced into C-and N-terminal regions of the DnaE N-and C-terminal segments (I N and I C ), respectively. A functional in vivo assay was initially used to examine the capability of these mutants to reconstitute an ALS in E. coli ER2744 cells. ALS catalyzes the first common step in the biosynthesis of branched-chain amino acids (26). E. coli strain ER2744, which lacks an active isoform II of ALS, is sensitive to feedback inhibition by valine, but its growth can be rescued by an active ALS expressed from a plasmid (27). When the gene encoding the ALS protein was split into two segments and each fragment was fused in frame to I N or I C , co-expression of the fusions was able to reconstitute an active ALS (23). Plate assays were conducted to examine the valine-sensitive growth of ER2744 cells co-transformed with N-and C-terminal fusion constructs, one possessing the wild-type intein segment and the other carrying a mutated segment. As shown in Fig. 4A, co-expression of the wild-type ALS intein fusions, ALS(N)-I N and I C -ALS(C), permitted the growth of the host cells, whereas coexpression of the non-fusion ALS fragments failed to rescue the growth of the host cells. The deletions in the N-terminal region of I C exhibited dramatic effects on the growth of the ER2744 host cells in the presence of valine. Removal of more than 3 residues, following the translation initiation methionine of I C , resulted in complete inhibition of growth of the host cells,  (3,4,5,8,9). Amino acid residues are numbered forward from the N-terminal splice junction. The conserved intein sequence blocks (A, B, F, and G) are labeled above the sequences. The secondary structure assignments for ␤ strands (␤1-␤10) and ␣ helices (␣1-␣2) for the Mxe GyrA intein are underlined. The endonuclease domain (256 amino acids) and DNA recognition region (DRR, 42 amino acids) of the Sce VMA intein were not included in the diagram. The mutated amino acid residues are boxed. B, a model structure of the Ssp DnaE intein based on the x-ray structure of the Mxe GyrA intein (9).
Ribbon drawing shows the region that corresponds to the C-terminal segment of DnaE intein (red) and the region that corresponds to the N-terminal segment of DnaE intein (green). The C-terminal 23 amino acid residues of I N , corresponding to the flexible linker region of the Mxe GyrA intein, is not shown. N and C termini of the intein are labeled N and C, respectively. Also labeled are the predicted amino acid sequences from residue 95 to 100 (I N ⌬23 to I N ⌬29; overlapping ␤9 strand) in white and the residues from residue 127 to130 (I C ⌬2 to I C ⌬6; overlapping ␤10 strand) in blue.
suggesting that an active ALS was not produced. In contrast, growth of the host cells expressing a mutated I C carrying 1 or 2 amino acid deletion indicated the production of a functional ALS. A deletion of 3 residues from I C resulted in partial growth. On the other hand, removal of 3-23 residues from the C terminus of I N rescued the cells from growth inhibition of valine (Fig. 4B), indicating that an active ALS was formed. Mutants carrying a deletion of 29 or 45 residues from I N exhibited partial or no growth. Furthermore, site-directed mutagenesis was performed to study the effect of amino acid substitutions in the highly conserved intein blocks B in I N and F in I C on the production of an active ALS. The residues Thr 69 and His 72 in block B and Phe 139 , Asp 140 , His 147 , Asn 148 , and Phe 149 in block F were substituted with Ala. For all these mutants, co-expression with the wild-type fusion protein permitted the growth of ER2744 cells in the presence of valine (Fig. 4C).
Analysis of Splicing Activities by Immunoblotting-Reconstitution of functional ALS could be due to either protein transsplicing or association between the two intein segments. Western blot analysis was carried out to detect if a spliced ALS product (59 kDa) was produced as a result of co-expression of both intein fusion constructs. As shown in Fig. 5, co-expression of the fusion proteins possessing the wild-type intein segments permitted protein trans-splicing to occur, generating a 59-kDa product, which reacted with antibodies against the ALS Nterminal fragment or the C-terminal fragment. A nonspecific 61-kDa protein cross-reacted with anti-ALS(N) antibody. The species corresponding to the full-length ALS product was also observed when 3-23 residues were deleted from the C terminus of I N (I N ⌬3-I N ⌬23) or 1 and 2 residues (I C ⌬1 and I C ⌬2) were deleted from the N-terminal region of I C (Fig. 5, A-C). These results are in agreement with the plate assay data showing that these mutants rescued the host cells from the growth inhibition by valine. A minor discrepancy was observed in the I C ⌬3 mutant, which exhibited splicing activity (Fig. 5C) but only partial growth in the plate assay (Fig. 4A); the reason for this difference is not clear. The spliced ALS product was not detected when a 29-or 45-amino acid deletion was introduced at the C terminus of I N or when 6 -12 residues were removed from the N-terminal region of I C . Furthermore, the substitutions in blocks B and F abolished trans-splicing, since no ALS product was detected using the anti-ALS(C) antibody (Fig. 5D), suggesting that these conserved residues are required for trans-splicing.
Detection of Intein Fragment Association by an in Vitro Assay-The observation that some intein mutants failed to mediate trans-splicing but rescued ER2744 host cells from growth inhibition by valine suggested that interaction of the two intein fragments may be sufficient for reconstitution of ALS activity without the formation of a spliced product. An in vitro assay was established to determine whether the mutated intein N or C fragments of the DnaE intein interact and if this interaction permitted trans-splicing in vitro. To facilitate the purification of the wild-type or mutated intein segments, the ALS N-terminal coding sequence was replaced by the DNA sequence encoding E. coli MBP (or M) (28) and the ALS C-terminal coding sequence was substituted with the DNA sequence encoding GST (or G) (25). The MBP-I N (MI N ) fusion proteins were purified and then added to the purified and glutathione resinbound I C -GST (I C G), as summarized in Table I. After a 45-min incubation at 4°C, the glutathione resin was washed thoroughly and samples of glutathione-Sepharose were taken and examined by SDS-PAGE. In the control experiment, MI N and I C G, possessing wild-type intein segments, interacted (Fig. 6A,  compare lanes 2 and 3) and mediated trans-splicing, producing MBP-GST (MG) (lanes 2-4). The chaperone protein DnaK, which co-purified with I C G, ran at a molecular mass of ϳ72 kDa on SDS-PAGE and was present in all the I C G samples. Alanine substitutions in blocks B and F abolished trans-splicing in vitro (Fig. 6), in agreement with Western blot analysis, which showed a lack of in vivo spliced product (Fig. 5D). These mutations, however, did not appear to affect the interaction between I N and I C since the MI N species remained bound to the I C G protein column after a thorough wash (Fig. 6, A and B,  lanes 2 and 3). N-terminal cleavage activity, as indicated by the appearance of a 42-kDa product, M, was detected in all the FIG. 4. Effect of various intein mutations on the growth inhibition of E. coli ER2744 by valine. ER2744 cells were co-transformed with two compatible plasmids expressing ALS-intein fusions, ALS(N)-I N and I C -ALS(C), one of which carries a mutated intein sequence as indicated. Plate assays were performed with ALS-intein fusions containing deletions of the N-terminal amino acids from I C (A), deletions of the C-terminal amino acids from I N (B), or amino acid substitutions in the conserved intein blocks B and F (C). Co-expression of both ALSintein fusion genes possessing the wild-type intein sequence rescued ER2744 host cells from growth inhibition by valine. In the controls, ER2744 cells were also transformed with plasmids expressing either wild-type ALS (wtALS) or ALS carrying V26A mutation (ALS), or both ALS(N) and ALS(C) fragments. Cells were plated on M9 agar medium supplemented with 100 g/ml valine and 0.3 mM IPTG and incubated at 30°C for 48 h. block B mutants except for the F139A mutant (Fig. 6B). Interestingly, the T69A or H72A mutants permitted DTT-induced cleavage at the N-terminal splice site (Fig. 6A, middle and right   panels, lane 4). The N148A mutant exhibited splicing activity at 30°C but not at 4°C (Fig. 6B, compare lanes 3 and 4), as confirmed by immunoblotting with antibodies against MBP or GST (data not shown). The data suggest that the association of the intein fragments account for the in vivo ALS activity observed in the block B and F mutants. The removal of only 3 residues from the N-terminal region of I C had no effect on interaction between I N and I C or on splicing and generated the spliced product, MG (Fig. 6C). However, deletion of a 10-amino acid region of I C abolished trans-splicing as well as its association with I N , as indicated by the absence of the spliced product MG (Fig. 6C, I C ⌬10, lane 2) and a dramatic decrease in the amount of MI N following the wash step (lanes 3 and 4). Similar results were observed when 6 or 8 residues were removed from I C (data not shown). The data also reveal that deletions of up to 29 residues from the C-terminal region of I N still allowed the association of I N and I C , whereas the 45-residue deletion in I N reduced the intein fragment association (Table I and Fig. 6C).

DISCUSSION
The work presented demonstrates that zinc inhibits transsplicing and trans-cleavage activities of a split intein encoded by the dnaE gene of Synechocystis sp. PCC6803, possibly by binding to Cys 160 , an essential catalytic residue at the C-terminal splice junction, and therefore may be a modulatory factor for protein splicing in vivo. Furthermore, the biochemical and molecular genetics experiments led to the identification of novel regions and structural determinants required for association and trans-splicing of the Ssp DnaE intein.
Inhibition of Protein trans-Splicing by Zinc Ion-Since the discovery of protein splicing a decade ago (29,31), no external energy source, such as ATP or a trans-acting factor, has been found to affect the process. Poland et al. (19) reported the discovery of a zinc atom at the catalytic center of the Sce VMA Ϫ ϩ Ϫ ϩ a Assayed with a mutated and a wild-type intein segment except for the control. Truncations were introduced at the C-terminal or N-terminal region of I N and I C , respectively.
b Assayed with E. coli ER2744 co-transformed with two compatible plasmids carrying the split E. coli ALS gene fused to the N-and C-terminal intein segments (Figs. 4 and 5). c Ϫ, absent; Ϫ/ϩ, weak; ϩ, strong; ND, not determined. d As determined by immunoblot analysis (Fig. 5). e As determined by the binding of MBP-I N to matrix-bound I C -GST (Fig. 6).
f Wild-type.
intein precursor, which possesses mutations to block its splicing activity. In this context, Cys 455 , His 453 , and Glu 80 , along with a water molecule were found to chelate a zinc atom with the coordinating angles and bond distances exhibiting characteristics of a structural zinc. The first amino acid of the Cextein, Cys 455 , participates in catalyzing splicing by transesterification (10). Numerous studies have shown that mutation of Cys 455 inhibits splicing of the Sce VMA intein, whether in the native host protein or in a foreign protein (10 -12). The identification of a single zinc atom per crystallized protein precursor molecule suggested that zinc binding by the intein is specific (19). We hypothesized that the chelation of a zinc atom would neutralize the nucleophile present at the C-terminal splice junction for its nucleophilic attack at the thioester bond formed at the N-terminal splice junction. The data presented here show that trans-splicing activity of the Ssp DnaE intein can be inhibited in vitro in the presence of 1-100 M zinc ion, which is ϳ2-200-fold in excess of protein concentration (Fig.  2B). As expected, the presence of EDTA, a divalent metal chelating agent, reversed the effect of zinc inhibition. The inhibition appears to be highly specific since other divalent metal ions, such as Ca 2ϩ and Mg 2ϩ , showed no significant effect on splicing. It is possible that only the splicing-deficient Sce VMA intein precursor could trap and co-crystallize with zinc, since the wild-type precursor molecules fold and undergo processing rapidly (10 -12, 31). Furthermore, the Ssp DnaE intein possesses neither a His residue at the penultimate position nor a Glu residue at the equivalent position (Glu 80 ) in block B, which are involved in the coordination of zinc in the x-ray structure of the Sce VMA intein (19). It is likely that residues at different positions are involved in the coordination of a zinc atom in the Ssp DnaE intein, as proposed by Mills and Paulus (30). Further understanding of the mode of zinc binding relies on a precise crystal structure of the Ssp DnaE intein.

Inhibition of N-terminal Cleavage by Zinc Ion-
The work presented also demonstrates that the presence of zinc ion blocks cleavage of the peptide bond at the N-terminal splice site ( Fig. 2A). It is possible that chelation of a zinc atom by Cys 160 may exclude the binding of DTT, thereby inhibiting its nucleophilic attack on the thioester bond. On the other hand, inactivation of Cys 160 by zinc chelation could affect the acyl rearrangement involving Cys 1 , thereby shifting the equilibrium of amide and thioester to favor amide conformation (Fig. 1). This scenario is supported by the observation that cleavage at the N-terminal splice junction can be inhibited by zinc in the mutant carrying the N159A substitution, which blocked splicing but permitted N-terminal cleavage, presumably by hydrolysis of the thioester bond (Fig. 2C). Furthermore, substitution of Cys 160 with an Ala residue resulted in no N-terminal cleavage, with or without DTT (Fig. 2D). It has been observed previously that the Cys residue at the C-terminal splice junction of the M. tuberculosis RecA intein has a low apparent pK a , which would facilitate the occurrence of transesterification by attacking the thioester bond involving Cys 1 (32). Thus, Cys 160 appears to play an important role in driving the first splicing step and shifting the equilibrium of amide and thioester.
Mapping the Regions Required for Intein Association-Protein splicing is presumably facilitated by inter-or intramolecular recognition between the N-and C-terminal regions of an intein, which may be separated by an endonuclease domain of more than 300 amino acid residues. For cis-splicing inteins, the tertiary folding occurs via intramolecular interactions of the intein sequence elements. The presence of a distinct endonuclease domain or a flexible linker may facilitate the formation of substructures and the folding process. The reconstitution of the split Ssp DnaE intein, however, requires a high affinity intermolecular interaction (33). The work presented here shows that the association function of the two intein segments is separable from its catalytic activity. trans-Splicing activity, however, appears to correlate with the interaction between the two intein halves, signifying that the interaction of the two fragments is necessary for the formation of the correct tertiary structure required for splicing. The data suggest that the Nterminal region of I C , adjacent to the conserved block F and part of ␤10, is essential for intein association. Intein C-terminal fragments carrying 6 -12-amino acid deletions are not capable of complementing the growth defect of ER2744 host cells or splicing in vivo or in vitro (Table I and Figs. 4 -6). The substantially lower affinity of I C possessing a deletion, for the wild-type I N could be due to the disruption of the threestranded ␤-sheets (␤10:␤5:␤6). The C-terminal 23 residues of I N appear not to be required for either association or splicing, probably because they constitute a flexible linker region (Fig.  3A). For the Mxe GyrA intein, the linker region forms a disordered loop and two ␣-helices extended from the ␤-core and does not appear to interact with the active site (9). It has yet to be determined whether the DnaE intein possesses a linker region with substructures corresponding to the ␣-helices of the GyrA intein (9). However, the DnaE intein sequence upstream of this region appears to align with the amino acid residues of the GyrA intein that participate in formation of the ␤9 strand, which appears to interact with the intein C-terminal segment, and therefore may be important for the formation and stabili- zation of the active site. Indeed, in the ALS fusion constructs further truncations of this region, as shown in the I N ⌬45 mutant, abolished splicing and failed to produce active ALS. The reduced interaction in vitro between I N ⌬45 and wild-type I C suggests that the region involved in formation of the ␤7 and ␤8 strands is important for both splicing and intein association (Fig. 6C). The 29-residue deletion, presumably disrupting only part of ␤9 strand, inhibited splicing but had little effect on association in vitro; however, partial growth was observed in the plate assay. The intein splicing and association activities could be affected by the extein context, which may result in the discrepancy between the in vivo and the in vitro data. These ALS intein fusion proteins could have suboptimal interactions, whereas MBP and GST, are effective at promoting the solubility of proteins to which they fused and probably interact proficiently (34).
Effect of Mutations in Blocks B and F on Splicing and Association Activities-The data from the in vivo functional screen and in vitro analysis indicate that the conserved residues in blocks B and F are crucial for splicing but not for association of the two intein fragments. Substitution of Thr 69 or His 72 in block B with an alanine residue inhibited splicing but still permitted interaction between the two intein segments. These mutants failed to generate the spliced ALS proteins (Fig. 5D) but were capable of complementing the valine-sensitive growth defect of ER2744 strain (Fig. 4C), suggesting that active ALS can be reconstituted by the association of the two intein halves. The conserved Thr 69 and His 72 in block B are implicated to participate in the acyl rearrangement, based on their positions in the active site of the Mxe GyrA intein or the Sce VMA intein (9,19). In this study acyl rearrangement, however, does not appear to absolutely depend on Thr 69 or His 72 , since the T69A and H72A mutant proteins exhibited DTT-induced cleavage activity (Fig. 6). It has been shown previously that mutation of the conserved His in block B of the Sce VMA intein abolished splicing activity in E. coli, but the single mutant was still capable of producing an active VMA protein in yeast (18). Thus, single amino acid substitutions may not completely block splicing as interactions from other intein residues may be sufficient to catalyze the reaction.
Our study provides the first experimental evidence that block F region is essential for splicing, but not association (Table I and Figs. 5 and 6). Crystallographic study of the Sce VMA intein reveals that several residues, including Ile 434 and Phe 444 in block F, participate in the formation of a hydrophobic surface at the back of the intein C-terminal asparagine residue. Alanine replacement of DnaE intein Phe 139 and Phe 149 , corresponding to Ile 434 and Phe 444 of the Sce VMA intein (Fig. 3) inhibited splicing, supporting the hypothesis that they may participate in the cyclization of asparagine (19). Ser 179 of the Mxe GyrA intein and Thr 435 of the Sce VMA intein, at the equivalent position of the DnaE intein residue Asp 140 , have been implicated in assisting the cyclization of intein C-terminal asparagine (9,19). Although DnaE intein residue His 147 is highly conserved among inteins (20 out of 24 eubacterial inteins), its role has not been elucidated. However, the presence of a His, at the equivalent position, in the vicinity of the active pocket of the Sce VMA intein suggests that it may play an important role in splicing (19).
In conclusion, we have characterized several important aspects of protein splicing. Zinc inhibition, together with the reversibility of EDTA, could provide a novel approach to modulate protein splicing and peptide bond cleavage. The in vivo functional screen along with an in vitro assay helped to dissect the regions of the intein essential for splicing and interaction between the two intein halves. Furthermore, a functional enzyme can be produced either by protein trans-splicing or by the association of two intein fragments (23,35). Therefore, the methods demonstrated here should open up new avenues in the application of self-splicing inteins to express and modify proteins.