Identification of Amino Acid Residues Involved in the Activity of Phosphomannose Isomerase-Guanosine 5’-Diphospho-~-mannose Pyrophosphorylase A BIFUNCTIONAL ENZYME IN THE ALGINATE BIOSYNTHETIC PATHWAY OF PSEUDOMONAS AERUGINOSA*

Phosphomannose isomerase-guanosine 5‘-diphospho- D-mannose pyrophosphorylase (PMI-GMP), which is encoded by the algA gene, catalyzes two noncontiguous steps in the alginate biosynthetic pathway of Pseudomo- nas aeruginosa: the isomerization of D-fructose 6-phos-phate to D-mannose 6-phosphate and the synthesis of GDP-D-mannose and PPi from GTP and D-mannose 1-phosphate. Amino acids that are required for the GMP enzyme activity were identified through site-directed mutagenesis of the algA gene. Mutation of Lys-176 to arginine, glutamine, or glutamate produced an enzyme whose K , for D-mannose 1-phosphate was 4704,200-fold greater than that measured for the wild type enzyme. In addition, these mutant enzymes had a lower V,, for the GMP activity as compared with the wild type PMI-GMP. These results indicate that Lys-175 is primarily involved in the binding of the substrate D-mannose 1-phosphate, although it is likely that other residues are required for the specificity of binding. Mutation of k g 1 9 to gluta- mine, histidine, or leucine The column fractions containing the 52-kDa chymotryptic fragment were pooled, concentrated using a Centricon-10 filter (Amicon), and then digested a second time with chymotrypsin. The purified chymotryptic fragment was then assayed for PMI and GMP activity as described above. The amino-terminal amino acid sequence was determined by Dr. Ka-Leung Ngai at the University of Illinois at Urbana.

cous exopolysaccharide called alginate (2). Alginate secreted by mucoid I! aeruginosa is a linear copolymer composed of @-1,4linked D-mannuronic acid (partially 0-acetylated) and the C-5 epimer L-guluronic acid (3-5). Alginate presents a substantial mechanical obstruction in the CF-affected lung and appears to reduce both the efficiency of the cellular immune response toward and the effectiveness of antibiotic therapy against €? aeruginosa that infect the lungs of CF patients (1). Elimination of the alginate barrier would likely enhance treatment of chronic pulmonary infections by €? aeruginosa and improve the prognosis for CF patients. Thus, we are studying the biochemistry and genetics of the alginate pathway in I! aeruginosa in an effort to identify nontoxic inhibitors of alginate biosynthetic enzymes.
Mucoid strains of €? aeruginosa produce low levels of the alginate biosynthetic enzymes (Fig. l ) , whereas these enzymes are either absent or greatly reduced in nonmucoid strains (6,7). Many of the alginate biosynthetic genes have been cloned and sequenced (8)(9)(10)(11)(12)(13)(14)(15)). In addition, cloning alginate genes under control of a strong promoter (e.g. tac) has been used to produce sufficient amounts of their respective gene products for identification of enzyme function (9, 11-14, 16, 17). This strategy also facilitated the purification of PMI-GMP (161, a bifunctional enzyme that catalyzes the first and third steps of alginate biosynthesis, and GDP-D-mannose dehydrogenase (171, which catalyzes the fourth step o f the alginate pathway of I! aeruginosa (Fig. 1).
The purified PMI-GMP protein exhibits a number of characteristics which suggest that the PMI and GMP enzymatic activities may reside in separate catalytic domains (16). First, PMI activity (but not GMP activity) is inhibited by sulfhydryl reagents. Second, GMP activity is dependent on the presence of either Mg2+ or Mn2+, whereas the PMI reaction utilizes a variety of divalent metals with Co2+ giving maximal activity.
Third, the substrates and products of the GMP reaction do not inhibit the PMI enzyme activity and vice uersa. Fourth, there are large structural differences between the substrates for PMI and GMP as well as mechanistic differences between an isomerization reaction and the charging of sugar phosphate with a nucleotide. In this study, we used site-directed mutagenesis and limited proteolysis to identify regions of the aZgA-encoded protein necessary for the PMI and GMP enzyme activities.
Strains and Plasmids-The bacterial strains and plasmids used in this study are shown in Table I. I? aeruginosa 8821 is a n alginateproducing (Alg') strain isolated from the sputum of a CF patient and, like other mucoid I? aeruginosa strains, spontaneously reverts to the nonmucoid (Alg-) form (18). The stable Alg' strain 8830 was obtained by chemical mutagenesis of the spontaneous nonmucoid strain 8822. The algA mutant strain 8853 was isolated aRer further mutagenesis of strain 8830 (18). The plasmid pDSl (Fig. 21, which was used for sitedirected mutagenesis of the algA gene, was constructed by cloning a 2-kilobase BamHI fragment, containing the algA gene, from PES119 (12) into pUC119 using standard methods (19).
Site-directed Mutagenesis ofalgA-Mutations in the algA gene were obtained using the oligonucleotide-directed in vitro mutagenesis system of Amersham (version 2) except that pUC119 was used as the cloning vector. Single-stranded DNA of the pUC119 aZgA derivative, pDS1, was obtained using the helper phage M13K07 (Promega). Mutant oligonucleotide primers were prepared by Operon Technology, Inc. (Alameda, CA). Mutations in the algA gene were confirmed by DNA sequencing using the dideoxynucleotide chain termination method (20) with the following modifications. Sequenase DNApolymerase (U. S. Biochemical Corp.) was used at 37 "C according to the directions of the manufacturer, and 7-deaza-dGTP was substituted for dGTP for all sequencing reactions to reduce compression artifacts (21). Preparation of sequencing gels, conditions of electrophoresis, and autoradiography were performed as described previously (22).
Each of the algA mutant genes was cloned as a 2-kilobase BamHI fragment into the BamHI site of the broad host range expression vector pMMB66HE using standard molecular biology techniques (19). The recombinant expression plasmids (Table I) were then mated into the algA mutant strain 8853 by triparental mating (23). Transconjugants were selected for on Pseudomonas isolation agar (Difco) containing 450 pg/ml carbenicillin. The transconjugants were then screened for the ability to produce alginate (i.e. complement strain 8853) by adding 1 m~ IFTG to the medium and observing colonies for a mucoid phenotype.
Purification of PMI-GMP Wild IJpe and Mutant Enzymes-The I? aeruginosa 8853 transconjugants were grown in 500 ml of Luna broth containing 450 pg/ml carbenicillin and 1 m~ IPTG at 37 "C with vigorous shaking (250 rpm). After an 8-h growth period, the cells were harvested by centrifugation a t 13,000 x g for 15 min, washed once with 0.9% NaCI, and resuspended in 16 ml of lysis buffer (100 m~ Tris-HCI, pH 7.0, 10 m~ MgCI,, 15% glycerol, 2 n" dithiothreitol). The cells were disrupted by sonic vibration in an oscillator (Branson) and centrifuged at 40,000 x g for 15 min. The resulting supernatants were then centrifuged at 100,000 x g for 1 h, and the clarified supernatants were used for PMI-GMP purification. PMI-GMP was purified using a Q-Sepharose column and a Bio-Gel TSK phenyl-5-PW HPLC column according to procedures published previously (16) except that MOPS buffer was replaced by 100 m~ Tris-HC1, pH 7.0. Fractions containing PMI-GMP activity were assessed for purity by SDS-PAGE and staining with Coomassie Blue G-250 (24). The wild type and mutant proteins were estimated to be greater than 95% pure.
PMI and GMP Assays-PMI activity was assayed in the reverse direction by the method of Slein (25) using a modified assay mixture (16). The rate of NADP' reduction was monitored at 340 nm at 25 "C in a Gilford Response I1 spectrophotometer. GMP activity was assayed in the forward direction by coupling the reaction to GDP-o-mannose dehydrogenase as described by Shinabarger et al. (16). The rate of NAD' reduction was monitored at 340 nm at 25 "C.
Proteolysis Studies-Purified PMI-GMP (5 mg) was incubated with 250 pg of chymotrypsin for 45 min at 25 "C in the presence of 1 m~ D-mannose 1-phosphate, 1 m~ GTP, and 10 mM MgCl,. The GMP substrates were found to increase greatly the stability of the major chymotryptic fragment. The chymotrypsin reaction was then treated with 250 pg of trypsidchymotrypsin inhibitor and subsequently diluted with 2 volumes of 100 n" Tris-HC1, pH 7.0, containing 10 m~ MgCl,, 2 m M dithiothreitol, and 15% glycerol. The sample was loaded onto a Mono-& column (Pharmacia) that had been equilibrated in the same buffer. Protein was eluted with a 60-ml linear gradient of C 2 5 0 m~ NaCl a t a flow rate of 1 ml/min. The column fractions containing the 52-kDa chymotryptic fragment were pooled, concentrated using a Centricon-10 filter (Amicon), and then digested a second time with chymotrypsin. The purified chymotryptic fragment was then assayed for PMI and GMP activity as described above. The amino-terminal amino acid sequence was determined by Dr. Ka-Leung Ngai at the University of Illinois at Urbana.

RESULTS AND DISCUSSION
Amino Acid Sequence Comparisons of PMI-GMP with Related Proteins-The amino acid sequence of the algA-encoded PMI-GMP was first compared with the sequences of proteins that catalyze similar enzymatic reactions to identify residues that are potentially important for PMI and/or GMP activity. No significant areas of homology were identified in a comparison of PMI-GMP with the PMI enzymes of Escherichia coli (261, Salmonella typhirnurium (271, or Rhizobium meliloti (28). PMI-GMP was, however, found to share a high degree of amino acid sequence similarity with the GMP enzymes RfbM and CpsB of S. typhimurium and E. coli ( Fig. 3; 29-32). The XanB protein of Xunthamonas campestris (331, a bifunctional PMI-GMP enzyme involved in xanthan gum synthesis, was also found to be highly related to PMI-GMP of the alginate biosynthetic pathway (Fig. 3).
A comparison of PMI-GMP with other isomerases and pyrophosphorylases revealed a weak homology (<23% identity) with the bacterial enzymes CDP-glucose pyrophosphorylase (29), UDP-glucose pyrophosphorylase (29, 34, 35) and ADP-glucose pyrophosphorylase (36, 37). Two regions of PMI-GMP were, however, found to be quite similar to ADP-glucose pyrophosphorylase. First, the Lys-175 region (FVEKF') of PMI-GMP is identical to the substrate binding site of bacterial ADP-glucose pyrophosphorylase (GlgC) and is highly conserved among the AlgA (PMI-GMP), RfbM, CpsB, and XanB proteins (Fig. 3). A similar amino acid sequence is present in the CDP-glucose pyrophosphorylase (29) and UDP-glucose pyrophosphorylase enzymes (29,341, and TDP-glucose pyrophosphorylase (29) also has remnants of this substrate binding site. This suggests that the amino acid sequence FVEKF' may be a substrate binding motif for this class of pyrophosphorylases. Second, the Lys-20 region of PMI-GMP is similar to the allosteric site of ADPglucose pyrophosphorylase. Nonetheless, Lys-20, the putative allosteric residue according to the homology with ADP-glucose pyrophosphorylase, is not conserved among the GMP class of proteins (Fig. 3).
Oligonucleotide-directed in Vitro Mutagenesis of the algA Gene-Fifteen independent mutations were made in the algA gene to change amino acids within the highly conserved regions of PMI-GMP (Fig. 3). The mutant genes were then cloned under control of the IPTG inducible tac promoter of the vector pMMB66HE (Table I), after which the recombinant plasmids were mated into the €? aeruginosa algA mutant 8853. SDS-PAGE analysis of cell-free extracts showed that all of the 8853 recombinant strains except for 8853[pSSTOP276] had a polypeptide that comigrated with the wild type PMI-GMP (data not shown). The algA gene in pSSTOP276 led to polypeptide chain termination because of conversion of the Ser-276 codon TCG to the stop codon TAG. PMI-GMP was not detected in cell-free extracts of 8853[pSSTOP276] presumably because the trun- cated polypeptide is unstable and is subject to degradation. Plasmid pSSTOP276 was not examined further. Plasmid pWT1, containing the wild type algA gene, complemented the Alg-defect of strain 8853 in the presence or absence of IPTG because the tac promoter is leaky in P aeruginosa.
Transconjugants containing eight of the mutant plasmids conferred a different phenotype to strain 8853, indicating that the PMI and/or GMP enzyme activities were severely disrupted by these aZgA mutations (Table 11). We purified and characterized PMI-GMP produced by these mutant plasmids. All of these algA mutations affected amino acids within regions of PMI-GMP which were homologous to the allosteric or substrate binding sites ofADP-glucose pyrophosphorylase ( Fig. 3; 36-38).

Kinetic Characterization of the PMI-GMP Mutant Enzymes-
Mutation of Lys-175 to arginine, glutamine, or glutamate reduced the v, , , for the GMP reaction compared with that of the wild type enzyme, where the V,,, of the mutant was found to decrease with increasingly acidic amino acid replacements (Table 11). These amino acid substitutions also increased the K , for o-mannose 1-phosphate 470-3,200-fold over that observed for wild type PMI-GMP. PMI activity, however, was relatively unaffected by these mutational changes. These results indicate that the primary role of Lys-175 is to bind the GMP substrate D-mannose 1-phosphate.
Hill et aZ. (38) found that mutational replacement of Lys-195 affected the kinetics of the E. coli ADP-glucose pyrophosphorylase reaction in a manner similar to that of the PMI-GMP Lys-175 mutations. The similarity to ADP-glucose pyrophosphorylase suggests that the PMI-GMP Lys-175 mutants might also have a higher K,,, for GDP-D-mannose, although the kinetics of the GMP reaction were not determined in the direction of pyrophosphorolysis. ADP-glucose pyrophosphorylase is thought to bind its substrate D-glucose 1-phosphate via an ionic interaction of the c-amino group of lysine with the negatively charged phosphate rather than with a hydroxyl group of the sugar (38). Although the size, shape, and charge of the amino acid are also important for proper substrate binding, the observed increase in the K , for D-glucose 1-phosphate as Lys-195 was replaced with increasingly acidic residues supports the view that lysine binding to phosphate is the basis for optimal binding affinity. The importance of Lys-175 of PMI-GMP and Lys-195 ofADP-glucose pyrophosphorylase for phosphate binding is further supported by the occurrence of the same phosphate binding sequence ( F V E K P ) in several other pyrophosphorylase enzymes. Thus, it is not surprising that ADP-glucose pyrophosphorylase and PMI-GMP share a common binding sequence for the phosphate group yet bind different sugar phosphates. Additional amino acid residues are, however, likely required for the specificity of binding that is observed for both these enzymes (16,38).
We were surprised that the K, for D-mannose 1-phosphate  did not consistently increase as Lys-175 was changed from a basic to a neutral to an acidic amino acid, especially since D-glucose 1-phosphate binding by the ADP-glucose pyrophosphorylase Lys-195 mutants followed this trend (38). ADP-glucose pyrophosphorylase has a preponderance of proline and basic amino acids downstream of Lys-195 which have been implicated as part of an exposed substrate binding loop (37,39,40). PMI-GMP, however, contains fewer prolines and basic amino acids outside the immediate substrate binding site. Thus, it seems that the microenvironment of the phosphate binding sites of PMI-GMP and ADP-glucose pyrophosphorylase could be different. This difference may allow glutamate to function (albeit less efficiently than lysine) for GMP activity, whereas the phosphate binding pocket of ADP-glucose pyrophosphorylase has a more stringent requirement for a positively charged amino acid. It is possible that the divalent metal cofactor normally coordinates with the phosphate group of the sugar to allow efficient binding to the e-amino group of the lysine (and to a lesser extent with an arginine residue). However, the negative charge of glutamate could also be neutralized by metal coordination via water groups. Perhaps, unlike ADPglucose pyrophosphorylase, the phosphate pocket of PMI-GMP supports metal coordination with glutamate. Replacing lysine with a n uncharged amino acid (e.g. glutamine) abolishes phosphate binding, leading to a higher K, for the sugar phosphate substrate of both proteins. Although we favor this hypothesis, we cannot entirely rule out that conformational changes account for a t least some of the differences between the wild type and Lys-175 mutant proteins. Determination of the crystal structure will allow us to examine further the microenvironment of the phosphate binding site, the effect of the mutations on the binding pocket, and the potential role of metal coordination in phosphate binding. Nonetheless, Lys-175 clearly participates in the binding of o-mannose 1-phosphate.
The efficiency of GTP binding was also slightly reduced (2-%fold) by replacing Lys-175 with arginine, glutamine, or glutamate (Table 11). It is quite possible that mutation of Lys-175 changes the conformation of D-mannose 1-phosphate binding which in turn affects the binding of GTP. Alternatively, this result may indicate that Lys-175 participates in the binding of GTP to the guanine binding site. PMI-GMP lacks a consensus GTP binding site (41), so it is unclear exactly which amino acids are involved in guanine binding. In contrast to PMI-GMP, the Lys-195 mutants of ADP-glucose pyrophosphorylase exhibit wild type levels ofATP binding (38). Tyr-114 has been shown to be important for the binding of the adenine moiety ofAMP, ATP, and ADP-glucose (40,4244). Although Lys-195 also binds AMP and ADP-glucose, ATP binds Tyr-114 in a manner that is thought to preclude an interaction of the a-phosphate oxygen with . If Lys-175 also has a role in GTP binding, then the guanine binding site of PMI-GMP would likely be oriented such that the a-phosphate oxygen of GTP can interact with Lys-175. The use of azido-GTP should prove useful in identifying the guanine binding site and in elucidating any interaction between Lys-175 and this site.
Mutation of Lys-20 to glutamine produced a protein that was unable to support alginate synthesis, yet kinetic analysis of the mutant enzyme failed to show any obvious reason as to why the protein is less efficient than the wild type enzyme (Table 11). It is possible that the Lys-20 change affects the forward PMI reaction by increasing the K, for o-fructose 6-phosphate, but we are unfortunately unable to measure accurately the kinetic constants for the D-fructose 6-phosphate to D-mannose 6-phosphate isomerization reaction. However, mutation of the neighboring Arg-19 to histidine, lysine, or leucine had a dramatic effect on GMP enzyme activity. These mutations increased the K,,, for GTP 4"7-fold, increased the K, for D-mannose l-phosphate 2-8-fold, and decreased the V,,, for the GMP reaction 2-fold as compared with the wild type enzyme (Table 11). PMI activity was relatively unaffected by these k g -1 9 mutations. Gardiol and Preiss (45) found that Lys-39 mutants of ADPglucose pyrophosphorylase behaved similarly to the Arg-19 PMI-GMP mutants: a lower apparent affinity for the nucleotide triphosphate (4-fold) and a reduced binding affinity for the sugar phosphate (6-fold) in the presence of the activators. This suggests that Arg-19 of PMI-GMP may play a role in the allosteric regulation of GMP activity, but it should be noted that we have never observed an allosteric effect for PMI-GMP in vitro. However, it is plausible that the allosteric effectods) is present in high enough concentrations in our assay system (e.g. one of the intermediates of the coupling enzymes or a contaminant of one of the substrates) so that the wild type PMI-GMP is fully activated in vitro. Alternatively, there is emerging evidence that activator binding by ADP-glucose pyrophosphorylase brings the glycine-rich region (analogous to the nucleotide binding P-loop motif) closer to Tyr-114 for more efficient binding of the nucleotide triphosphate (46). Thus, Arg-19 may be part of the GTP binding site involving glycine residues near Ser-10 (Fig. 3). Indeed, there seems to be a big effect of mutating k g -1 9 on the K,,, of GTP. This site in PMI-GMP may not PMI-GMP purification was described under "Experimental Procedures." PMI activity was measured in the reverse direction, and GMP activity

Characteristics of mutant PMZ-GMP Droteins
The K175Q mutant no longer shows the preference for colbalt over magnesium for the PMI activity which is observed for the wild type producing (M), or nonmucoid (N).
was measured in the forward direction relative to alginate synthesis (16).
PMI-GMP and the rest of the mutant proteins.
require a n allosteric activator for function. Since Haugen and Preiss (47) found that ADP-glucose pyrophosphorylase first binds ATP and then binds D-glucose 1-phosphate, it follows that D-mannose 1-phosphate binding may also be reduced for the kg-19 mutants if the binding of GTP is required prior to binding of the sugar phosphate.
The ADP-glucose pyrophosphorylase allosteric site is highly basic, and at least 1 other lysine residue and 1 arginine residue have been implicated in the allosteric regulation of this enzyme (39,48,49). It is thought that the cationic groups together effectively reduce the pK of the €-amino group of Lys-39 and constitute the anion binding site (48). For PMI-GMP, k g -1 9 is a n important residue for GMP activity, and it is possible that several neighboring residues (e.g. k g -1 3 ,  contribute to the basic nature of this region. Guanidium groups have been shown to bind phosphate (501, and there is a precedent for the involvement of arginyl residues in the binding sites of anionic substrates and cofactors (51)(52)(53). Thus, k g -1 9 and the surrounding region have the characteristics of an anion binding site that, perhaps, is involved in binding of GTP a n d o r an allosteric effector.
Plasmid pSA12 required IPTG induction to restore strain 8853 to Alg', indicating that the Ser-12 to alanine replacement did not completely abolish the PMI and GMP activities of the Ala-12 mutant enzyme. Table I1 shows that the Ala-12 mutant protein had V , , , values for the PMI and GMP reactions which were 15 and 44%, respectively, of the wild type levels. However, the K , measured for D-mannose 6-phosphate, D-mannose 1-phosphate, and GTP were actually lower than that found for the wild type protein. It appears that IPTG induction can compensate for the loss in the reaction rate of the Ala-12 mutant enzyme, whereas IPTG induction does not compensate for the decreased substrate binding affinity (i.e. higher K,) that was observed for the other mutant proteins. The high glycine content in the region of PMI-GMP near Ser-12 suggest a structural role for this area of the protein. Therefore, it is likely that mutation of Ser-12 to alanine disrupts a n important structure required both for PMI and GMP enzyme activities. Replacing Val-321 with methionine has a similar effect on PMI-GMP where Val-321 was also postulated to play a structural role (16). Investigation of Amino Acid Residues Required for PMI Activity-Although the comparison of PMI-GMP with related proteins was useful in identifying amino acids that are important for GMP enzyme activity, none of the site-directed mutants selectively affected PMI activity. Therefore, a different approach was taken t o identify amino acids that are important for PMI activity. Wild type PMI-GMP was incubated with a variety of proteolytic enzymes in an effort to identify regions of the protein which are involved in PMI enzymatic activity. Chymotrypsin digestion of PMI-GMP produced a major proteolytic fragment of 52 kDa which had low PMI activity but normal levels of GMP activity. In addition, the chymotryptic fragment could be selectively protected from further digestion by coincubation with the GMP substrates. Thus, we purified the 52-kDa chymotryptic fragment to assure that intact PMI-GMP and other digestion products were not contributing to the enzymatic activities (Fig. 4). The 52-kDa chymotryptic fragment only retained 8% of the PMI activity of the undigested protein but retained 81% of the GMP activity of the intact protein. The amino-terminal amino acid sequence of the 52-kDa chymotryptic fragment (MIPVILSGGT) was found to be 90% identical to the undigested 53-kDa PMI-GMP, indicating that chymotrypsin cleaves about 1 kDa from the carboxyl terminus of PMI-GMP. Chymotrypsin most likely cuts at Leu-472 or Tyr-465. These results indicate that the carboxyl-terminal region of the protein is important for PMI activity as loss of the carboxyl terminus results in low PMI activity. Koplin et al. (33) arrived at the same conclusion using insertion mutants of the xanB gene and assaying the gene product for PMI and GMP activity. One of the insertions affected a region of the XanB protein which corresponded to the chymotryptic cut site of the algAencoded PMI-GMP. This mutant protein had lost PMI activity while maintaining 40% of the wild type GMP activity. GMP activity was still detectable, albeit in low amounts, even when the last 113 amino acids of the XanB protein had been disrupted by insertion into the xanB gene (33).
In Summary-The major goal of this study was to define the regions of PMI-GMP responsible for each of the enzymatic activities. I t is apparent that the amino-terminal half of the protein contains a t least two regions critical for GMP enzyme activity. Lys-175 is involved in the binding of D-mannose 1-phosphate, presumably via interaction of the phosphate group with the €-amino group of Lys-175, a n d k g -1 9 probably forms part of the GTP binding site. The homology of the aminoterminal half of PMI-GMP with several pyrophosphorylases adds further support for the idea that the amino-terminal portion is important for GMP activity. The chymotrypsin studies indicate that the carboxyl-terminal half of the protein is important for PMI activity. Thus far, it appears that PMI and GMP are in catalytically distinct domains. This is further SUPported by secondary structure predictions which suggest that the amino-terminal half of PMI-GMP is composed of mostly cy-helix, whereas the carboxyl-terminal half of the protein consists of an d/P-domain (data not shown). However, the results also show that some parts of the protein (ie. Ser-12 and Val-321) are important for both PMI and GMP activities. Identifi-