Biosynthesis of Proline

LASZLO N. CSONKA¹* AND THOMAS LEISINGER²

[SECTION EDITOR: GEORGES COHEN]

Posted August 8, 2007

Department of Biological Sciences, Purdue University, West Lafayette, IN 47907-1392,¹ and Institut für Mikrobiologie, Wolfgang Pauli Strasse 10, CH-8093 Zürich, Switzerland²

*Corresponding author. Phone: (765) 494-4969, Fax: (765) 496-1496, E-mail: This e-mail address is being protected from spambots. You need JavaScript enabled to view it

Although proline is commonly considered to be one of the "amino acids," this is a misnomer—it is an imino acid. Proline was among the last biosynthetic precursors to have its biosynthetic pathway unraveled. This was because of difficulties in developing sensitive and specific assays for the first two enzymes of the pathway. Proline is also unique among macromolecular building blocks in that the expression of its biosynthetic genes is not sensitive to the availability of the end product.

Vogel and Davis (117) first proposed that proline is derived from glutamate via glutamate γ-semialdehyde (GSA) and its spontaneous cyclization product Δ¹-pyrroline-5-carboxylate (P5C) (Fig. 1). This proposal was based on the observation that a certain proline auxotroph (now known to be a proC mutant) excreted P5C, which could support the growth of a second type of proline auxotroph (a proB or proA mutant). Radioactive tracing experiments showed that unlabeled GSA decreased the incorporation of L-[^l4C]glutamate into proline (118) and thus supported the proposed route. By analogy to the homoserine and arginine biosynthetic pathways, which involve the phosphorylation of a carboxylate group prior to its reduction (86a) (Chapter Biosynthesis of Arginine and Polyamines), it was suggested that the initial step in the proline biosynthetic sequence is likewise transfer of phosphate to glutamate, providing an efficient leaving group that drives the reduction of the γ-carboxyl group (118). The next stage in the unraveling of proline biosynthesis was the demonstration that proline production is subject to feedback regulation at some step prior to the formation of GSA (110). Baich and Pierson (8) provided evidence for a proline-inhibitable enzyme activity catalyzing the ATP-dependent phosphorylation of glutamic acid. The other two enzymes of the pathway, γ-glutamyl phosphate reductase (GPR) and P5C reductase (P5CR), were detected in crude extracts of Escherichia coli (7), and subsequently all three enzymes were confirmed in a number of other bacteria (12, 63, 64, 101, 111). Proline is synthesized by the pathway shown in Fig. 1 in all eubacteria, Saccharomyces cerevisiae, plants, and animals. However, in higher eukaryotes, there may also be a second route to P5C via ornithine (29). Orthologous genes for all three eubacterium-type proline biosynthetic enzymes can be discerned (with tblastn Expect value scores against the bacterial enzymes <10⁻⁷⁶) in the Archaea Natronomonas pharaonis, Methanococcoides, and Methanosarcina, but only P5CR, not the other two enzymes, can be recognized in Pyrococcus furiosus, Thermococcus kodakaraensis, Thermoplasma volcanium, Thermoplasma acidophilum, Aeropyrum pernix, Ferroplasma acidarmanus, Sulfolobus, Picrophilus torridus, and Methanothermobacter thermautotrophicus. However, no good matches to any of the three standard proline biosynthetic enzymes could be observed in a number of other archaebacteria, raising the possibilities that these organisms may use a different pathway (see below) or carry out the same reactions with very different enzymes.

Fig. 1Proline biosynthetic pathway.

A review by Adams and Frank (1) summarized the information available on the metabolism of proline and hydroxyproline in microbes and higher organisms as of 1979. The enzymology and genetics of proline biosynthesis in E. coli were last reviewed in the second edition of this compendium (68). Since that time, only modest new information on proline biosynthesis has accumulated, and interest in proline has shifted to its role as an osmoprotectant.

This chapter recapitulates the findings on the biosynthesis and transport of proline. The degradation of this imino acid is covered in this compendium in Chapter Catabolism of Amino Acids and Related Compounds, "Catabolism of Amino Acids and Related Compounds," by Larry Reitzer.

THE PATHWAY

γ-Glutamyl Kinase

γ-Glutamyl kinase (GK) catalyzes the ATP-dependent phosphorylation of L-glutamic acid. Studies of this reaction in crude extracts and the purification of the enzyme have been hampered by three experimental difficulties. First, the extreme lability of γ-glutamyl phosphate (GP) and its tendency to cyclize to 5-oxopyrrolidine-2-carboxylate (110) complicated the identification of the product of the first enzymatic step. However, the enzyme-dependent formation of γ-glutamyl hydroxamate and P_i from glutamate, ATP, Mg²⁺, and hydroxylamine (6) was compatible with GP being the first intermediate of the pathway. When GPR, which catalyzes the second reaction of the pathway, was examined with a homogeneous preparation in the reverse of its biosynthetic direction, the product was verified to be 5-oxopyrrolidine-2-carboxylic acid (46). Since the latter compound arises from GP (61), this observation is in accordance with the scheme in Fig. 1, in which GP is an activated intermediate between the first and the second biosynthetic steps (48). Considerations of the lability of free GP led to the notion that the compound exists in vivo only as an enzyme-bound intermediate. Work with the glutamate analog cis-cycloglutamate supports the formation of GP and suggests that this intermediate reacts with some moiety on GPR (such as a thiol) to form a γ-glutamyl-enzyme complex (103).

Second, the proline-specific GK catalyzes the same activation reaction of glutamate as glutamine synthetase. Although the two enzymes can be differentiated by the fact that GK is sensitive to feedback inhibition by proline, whereas glutamine synthetase responds to different effectors (6, 50), the presence of both enzymes in crude cell extracts makes the assay of the former imprecise. With the hydroxamate assay used initially, the background due to glutamine synthetase can be 10-fold higher than the activity due to the proline-specific GK in crude extracts of the wild-type E. coli and Salmonella enterica serovar Typhimurium. The large background can be eliminated by a glutamine synthetase (glnA) mutation (S. Kustu and L. N. Csonka, unpublished observations).

The third obstacle was that, for GK to have activity in the hydroxamate assay that was used in the early experiments, it must be associated with GPR (107). This association is not required when the activity of GK is assayed as glutamate-dependent formation of P_i from ATP, which relies on the spontaneous hydrolysis of GP to 5-oxopyrrolidine-2-carboxylate and P_i (103, 108). The existence of an enzyme complex ensuring the direct transfer of an unstable intermediate from the first to the second enzyme had been suggested on the basis of gel filtration studies with crude extracts, and it appeared that the complex is very labile in vitro (50). In this context, it is interesting that in plants and animals, the first two steps of proline biosynthesis are catalyzed by a bifunctional enzyme, P5C synthetase, which has domains that correspond to the ProB and ProA proteins of E. coli (52).

It is now clear that early attempts to purify GK were unsuccessful because they led to the dissociation of the enzyme complex, with a consequent loss of hydroxamate-forming activity of the first enzyme. Purification of GK from E. coli was facilitated by the expression of the proB and proA genes from a high-copy-number plasmid (31, 107) and the development of a specific coupled assay based on the NADPH-dependent reduction of GP by GPR. In this assay, a large excess of highly purified GPR was added to support the activity of GK. Initial gel filtration studies suggested that the GK holoenzmye contains six subunits (107), but this conclusion was revised on the basis of cross-linking studies that indicated that the enzyme is a homotetramer (87).

GK exhibits positive cooperativity with increasing glutamate concentration and has half-maximal activity with 15 to 40 mM concentrations of this substrate (103, 107). It has hyperbolic kinetics with respect to ATP with a K_m of ~2 mM. ADP is a competitive inhibitor of ATP binding, with an apparent K_i of approximately 0.7 mM (103). Sensitivity to inhibition by ADP may lead to a fine-tuning of proline biosynthesis by the energy charge. Among glutamate analogs, cis-cycloglutamate (cis-1-amino-1,3-dicarboxycyclohexane) is a good substrate for GK, displaying hyperbolic saturation kinetics and a K_m of 1 mM (103); α-methyl-D,L-glutamate and γ-methyl-D,L-glutamate also support low activity as substrates. D-Glutamate is neither a substrate nor an inhibitor (103).

Typical of many enzymes catalyzing the first committed step of pathways, GK is sensitive to end product inhibition. This result was first demonstrated with resting cells of a proC mutant of E. coli by comparing the rates of conversion of glutamate to GSA in the presence and absence of proline (8, 110). The sensitivity of GK to proline was directly confirmed with crude extracts (6, 50). The K_i of the E. coli enzyme for L-proline has been reported to be in the range of 7 μM to 0.09 mM in the presence of 50 mM L-glutamate (103, 107, 108). The glutamate concentration that supports 50% activity increases with increasing proline concentration, suggesting that proline may be a competitive inhibitor of glutamate binding (107). The toxic metabolite 3,4-dehydroproline is also an inhibitor of the enzyme, with approximately threefold lower affinity than proline (107, 115).

GK, aspartate kinase, and N-acetylglutamate kinase belong to the amino acid kinase family of proteins (accession PF00696; http://www.sanger.ac.uk/Software/Pfam), which also includes carbamate kinase, carbamoyl phosphate synthetase, and uridylate kinase. These enzymes show sequence conservation in their central region, which has been suggested to contribute to the structure of the catalytic site (39, 80). Site-directed mutagenesis was used to identify residues that constitute the active site of E. coli GK (87b). This analysis indicated that there is an overlap between the binding sites for glutamate and the allosteric inhibitor proline, suggesting that proline competes with the binding of glutamate. The crystal structure of the GK from Campylobacter jejuni has been recently determined and is available at Protein Data Bank as ID 2AKO.

The GKs of most microorganisms, including the 367-amino-acid enzyme from E. coli, have a C-terminal tail of ~100 amino acids that has sequence similarities to PUA domain RNA binding proteins (4). However, there are shorter GKs that lack this tail, such as the 251-amino-acid enzyme predicted from the genomic sequence of C. jejuni (accession number AL139074, nucleotides 105956 to 10607), the 267-amino-acid enzyme from Streptococcus thermophilus (70, 74), and the 271-amino-acid eubacterium-type GK cloned from a tomato cDNA library (39). The latter two have been shown to be enzymatically active and sensitive to feedback inhibition by proline, indicating that the tail is not necessary for catalytic function, assembly, or regulation of the enzyme. Deletions have been introduced into the E. coli proB gene that removed 35 residues (104) or 108 residues (87a) from the C-terminal part of GK. Deletion of the PUA domain decreased the Mg²⁺ requirement of the enzyme but did not alter its ability to complement proB null mutations, confirming that the tail is not necessary for the in vivo function of the enzyme. The GK tail shows low similarity to the N-terminal portion of the GPR domain of the bifunctional P5C synthetases from plants, raising the possibility that GK and GPR may have evolved from a common ancestor by gene duplication (52).

The E. coli proB⁺ gene can complement pro1 (GK) mutations in Saccharomyces cerevisiae (82), and conversely, the S. cerevisiae PRO1, PRO2, and PRO3 genes can complement proB, proA, and proC mutations, respectively, in E. coli or S. enterica serovar Typhimurium (113). These observations were not only instrumental in the identification of the proline biosynthetic genes in yeast but also suggested that, if it is necessary for GK and GPR to associate, a functional complex can be formed between the bacterial and fungal enzymes.

γ-Glutamyl Phosphate Reductase

GPR catalyzes the NADPH-dependent reduction of GP to GSA. Because GP is not stable, the GK activity has been measured in reverse of the in vivo direction, the NADP⁺ and phosphate-dependent oxidation of GSA. The actual compound supplied in this assay is P5C, which is in rapid equilibrium with the straight-chain GSA that probably serves as the true substrate of the enzyme. D,L-P5C can be prepared by periodate oxidation of δ-hydroxylysine (119, 123). An enzymatic method for the preparation of L-P5C from L-ornithine with a partially purified preparation of ornithine 5-aminotransferase from Pseudomonas aeruginosa has also been described (45).

GPR was purified from an E. coli strain expressing the proB⁺A⁺ genes from high-copy-number plasmids, and 1,200-fold purification yielded a homogeneous product (31, 46, 107, 108). Different values for the molecular mass of the native protein have been reported, but overall they suggest a hexameric subunit arrangement. Studies with the pure enzyme showed that the reaction is highly specific for NADP⁺ and P5C (D,L-mixture), but arsenate, molybdate, and tungstate can substitute for phosphate. The kinetics of the reaction are consistent with a rapid-equilibrium, random-order mechanism (47).

The proB and proA genes of E. coli were fused into a single open reading frame by a series of steps that involved changing the C-terminal arginine of GK to proline, removing the translation termination codon, and adding a linker of four amino acids between GK and GPR (76). The fusion product, expressed from a high-copy-number plasmid, could complement a proBA deletion, and it increased proline accumulation by about twofold.

The proline auxotrophy of proA mutants of E. coli can be corrected by high-level expression of the E. coli or Corynebacterium glutamicum asd gene, which specifies aspartate semialdehyde dehydrogenase (i.e., aspartyl phosphate reductase [86a]) (105). This result suggests that, contrary to the generally accepted view, it may not be necessary to pass GP directly from GK to GPR or that, if a complex is necessary, it can be formed adequately between GK and aspartate semialdehyde dehydrogenase.

Δ1-Pyrroline-5-Carboxylate Reductase

The final step in proline biosynthesis is catalyzed by P5CR. In E. coli crude extracts, this enzyme interferes with the assay of GPR because both enzymes use P5C as a substrate and the NADPH formed by GPR can be consumed as a substrate in the P5CR reaction. A method for the rapid removal of P5CR from cell extracts was devised, which facilitated the initial assignment of the gene-enzyme relationships in proA and proB mutants (49).

A partially purified preparation of P5CR that was sufficiently free of competing enzyme activities for kinetic studies was obtained from an E. coli strain that overproduced this enzyme at a 200-fold increased level (98). The enzyme exhibited Michaelis-Menten kinetics, with K_mvalues of 0.15 and 0.03 mM for D,L-P5C and NADPH, respectively. Proline and NADP⁺ act as competitive inhibitors of the forward reaction with approximate K_ivalues of 15.0 and 0.6 mM, respectively. P5CR was later purified to homogeneity from an E. coli strain synthesizing the enzyme at a 190-fold-increased level from plasmid pGW7 proC. The homogeneous preparation was used to determine the amino- and carboxy-terminal amino acid sequences but was not characterized with respect to its catalytic properties (32).

SECOND ROUTE TO GSA IN argD MUTANTS

The second and third reactions of the arginine biosynthetic pathway are analogous to the first and second steps of proline biosynthesis, except that the intermediates in the arginine pathway carry an N-acetyl modification, which is thought to block the cyclization of the glutamate semialdehyde moiety (Fig. 2) (Chapter Biosynthesis of Arginine and Polyamines). The similarity of the initial intermediates in the two pathways forms the basis for the phenotypic suppression of proB or proA mutations by N-acetylornithine aminotransferase (argD) mutations, which has been reported in E. coli (13, 54) and in S. enterica serovar Typhimurium (66).

Fig. 2Relationship of proline biosynthesis to the pathways for arginine biosynthesis and proline catabolism. argA to argE are the genes for the enzymes catalyzing the first five reactions of arginine biosynthesis. The putA gene specifies a bifunctional dehydrogenase that oxidizes proline to P5C (proline dehydrogenase) and P5C into glutamate (P5C dehydrogenase). FAD, flavin adenine dinucleotide; FADH2, reduced flavin adenine dinucleotide.

Strains carrying an argD mutation accumulate N-acetylglutamate γ-semialdehyde. Despite the fact that argD mutants lack one of the arginine biosynthetic enzymes, they are not arginine auxotrophs because nonspecific aminotransferase(s) can substitute for the missing enzyme. The N-acetylglutamate γ-semialdehyde that is accumulated to high levels in argD mutants can be converted into GSA by the argE gene product, N-acetylornithine deacetylase, the catalyst for the fifth step of arginine biosynthesis, and consequently, argD mutations can suppress proB or proA mutations (Fig. 2). However, proline biosynthesis in argD proB or argD proA double mutants no longer responds to feedback inhibition by proline but is now subject to regulation by arginine. Such mutants excrete proline on minimal medium and are resistant to the proline analogs 3,4-dehydroproline and azetidine-2-carboxylate (13). The relationship between arginine and proline biosynthesis can be exploited for the selection of regulatory mutations in the arginine pathway. The proB argD or proA argD double mutants grow rapidly in minimal medium in the absence of proline, but they are inhibited by arginine. The reason for the latter phenotype is that arginine represses the transcription of the arg regulon and inhibits the activity of N-acetylglutamate synthase (the first enzyme of the arginine pathway). It is possible to obtain mutant derivatives that grow slowly in minimal medium containing arginine (37); these strains now carry argR mutations, and therefore, they express the arg biosynthetic enzymes constitutively. The proBA argD argR mutants excrete proline in arginine-deficient media, but the excretion is blocked by supplementation with arginine because of feedback inhibition of N-acetylglutamate synthase. Cultivation of the proBA argD argR triple mutants in minimal medium with arginine enriches for more quickly growing quadruple mutants that excrete proline in the presence of arginine. These strains have acquired an additional argA mutation that renders N-acetylglutamate synthase less sensitive to feedback inhibition by arginine.

GENETICS OF PROLINE METABOLISM

Genes involved in the metabolism of proline in E. coli and Salmonella are summarized in Table 1.

Table 1Genes and gene products of proline metabolism in E. coli and Salmonella

Selection of Proline Auxotrophs

There are two unexplained observations that may facilitate the isolation of proline auxotrophs. It has been reported that proB and proA mutants are resistant to 4-nitropyridine, and this compound has been suggested to be useful for the positive selection of those two types of auxotrophs (53) (although this could not be replicated successfully with S. enterica serovar Typhimurium in the laboratory of L. N. Csonka). Resting cells of most auxotrophs are killed by penicillin, whereas resting cells of proline-requiring mutants are insensitive to the antibiotic (97). Preferential recovery of mutants lacking the three proline biosynthetic genes was observed upon penicillin treatment of a mutagenized culture.

COMPARATIVE GENOMICS OF THE PROLINE BIOSYNTHETIC GENES

Among the completed genomic sequences of Enterobacteriaceae, genes specifying all three proline biosynthetic enzymes can be discerned in E. coli, Shigella, S. enterica, Serratia marcescens, Erwinia carotovora, Yersinia, Photorhabdus luminescens, and Sodalis glossinidius strain 'morsitans.' However, there do not appear to be orthologs of any of these genes in Buchnera aphidicola and Candidatus blochmannia, and Wigglesworthia glossinidia has only proC and lacks proB and proA. The latter three enteric bacteria have small genomes of 0.62 to 0.79 Mbp and are obligate insect symbionts, and so they may be dependent on their hosts for P5C or proline.

In those enteric bacteria that have all three proline biosynthetic genes, proB and proA are adjacent to each other and are transcribed as a single operon but proC is separated from proBA by at least 29 kb. Comparison of the sequences of these genes reveals that proB and proA are much more conserved across species both in their amino acid and in their nucleotide sequences than proC is (Table 2). This result suggests that the three genes in this single pathway have undergone very different rates of mutation. One explanation may be that the need for GK and GPR to interact imposes a higher constraint on their amino acid sequences than is placed on that of P5CR. Another possibility may be that the proC genes in different organisms may have arisen from different paralogs by convergent evolution whereas the proBA genes descended from unique ancestors.

Table 2Comparison of the predicted amino acid and nucleotide sequences of the proline biosynthesis genes of Enterobacteriaceae

proBA and proC Promoters

Alignment of the nucleotide sequences upstream of the proB gene indicates that these regions are highly conserved across the eight enteric species (Fig. 3A), although they appear to fall into two subgroups in which the sequences show closer relatedness to one another than to sequences in the other subgroup. The two subgroups consist of the sequences of (i) E. coli, Shigella, Salmonella, Erwinia, and Photorhabdus luminescens and (ii) Serratia marcescens, Yersinia, and Sodalis glossinidius. The proB mRNA start site has been determined in vivo in Serratia marcescens and possible −35 (TTGGCA) and −10 (TACAAA) elements have been proposed for a σ⁷⁰ promoter (81). Except for Sodalis glossinidius, the −35 sequence is maintained identically in the other seven enteric bacteria, and possible −10 elements can be found 17 nucleotides after the putative −35 element with a variance of at most 2 bp (also TACAAA in Yersinia, TACAAC in Erwinia, and TAAAAC in E. coli, Shigella, and Salmonella). However, there does not seem to be a good match for −35 and −10 sequences in Sodalis glossinidius. It has been noted that there is an inverted repeat encompassing the −35 element in Serratia marcescens (80). Inverted repeats with different sequence compositions and slightly different positions can be found around the −35 element in other organisms. The fact that such structures are conserved in various species hints that they may be involved in the expression of the proBA operon, although this has not been addressed experimentally.

Fig. 3Promoter regions of the proline biosynthetic genes of enteric bacteria. The sequences used for this comparison correspond to the strains listed in Table 2. For each strain, the pairs of sequences highlighted with yellow, green, or pink backgrounds are inverted repeats. Asterisks indicate residues that are conserved across the species in each group. (A) Promoter regions of the proBA operon. The boxed sequences TTGGCA and TAAAAC, TACAAC, and TACAAA are proposed as the −35 and −10 components of the promoter (see the text). The boxed ATG sequence in green is the proB translation start site. The sequences from E. coli and Shigella flexneri are identical in the region shown. (B) Promoter regions of the proC gene. The boxed sequences TTGCCT and TATGCT are possible −35 and −10 elements of the proC promoter in E. coli and Shigella flexneri (see the text). (i) For E. coli, Shigella flexneri, and S. enterica serovar Typhimurium, the boxed sequence CAT in green shows the translation start site of the neighboring gene yaiI, read in the opposite direction from proC. For Erwinia carotovora and Yersinia pestis (ii), Serratia marcescens, Photorhabdus luminescens, and Sodalis glossinidius (iii), and Wigglesworthia glossinidia (iv), the boxed sequences TAA, TGA, and TAG in red are the translation stop codons for the upstream yggS gene.

Source: The inverted repeat sequences in Fig. 3 and 4 were identified with the program Stemmer. We acknowledge P. T. Gilham (Department of Biological Sciences, Purdue University) for help with this algorithm.

There have been no biochemical studies of the transcription of the proBA genes, but the following data suggest that they are read as an operon from proB to proA. Complementation tests with S. enterica serovar Typhimurium demonstrated that insertions in proB are polar on proA (73). Omori et al. (80) isolated a single-base-pair substitution at position −16 in the putative proB promoter in Serratia marcescens that caused a fourfold increase in the production of both GK and GPR proteins, corroborating the inference that proB and proA are read from a common promoter. There is no indication for a separate terminator for proB, whereas possible Rho-dependent transcription terminators have been suggested to occur after the proA gene (see below). In all of the enteric organisms mentioned above, proB is separated from proA by 9 to 11 nucleotides that contain a predicted ribosome binding site (data not shown).

In accord with the lower level of conservation of the proC coding sequences, the transcriptional start regions of this gene also show much lower cross-species preservation than those of the proBA genes (Fig. 3B). The regions upstream of the proC gene can be assigned into four subgroups on the basis of their similarity, with the subgroups consisting of sequences from (i) E. coli, Shigella, and Salmonella; (ii) Erwinia and Yersinia; (iii) Serratia marcescens, Photorhabdus luminescens, and Sodalis glossinidius; and (iv) Wigglesworthia glossinidia (which has proC but not proBA).

In E. coli and Shigella, the sequence TTGCCT 17N TATGC located 48 nucleotides before the proC gene (Fig. 3B) matches sufficiently with sequences of known promoters (71) to hint that it may function in this capacity. The hypothetical 35 element TTGCCT is also present in Salmonella, but there is no good match for a counterpart to the −10 element in this organism. There do not appear to be obvious conserved promoter elements for the proC gene in Erwinia, Yersinia, Photorhabdus luminescens, Sodalis glossinidius, and Wigglesworthia glossinidia. Immediately upstream of proC in of these five organisms, there is a conserved gene that is orthologous to the E. coli yggS (putative alanine racemase) gene. The lack of a recognizable terminator at the end of the yggS orthologs and the short distance of 19 to 37 nucleotides between this gene and proC raise the possibility that the promoter for proC may be within or in front of the yggS gene in these organisms.

Comparison of the sequences upstream of proBA and proC to each other in each of the species did not reveal any obvious similar themes, which may be indicative of some common transcriptional regulatory mechanism.

proBA and proC Terminators

The sequences downstream of the proBA operon, which would be expected to contain the transcriptional terminators, show less conservation across species than the coding sequences or promoters (Fig. 4A). These sequences can be assigned into two subsets exhibiting closer similarities: sequences from (i) E. coli, Shigella, Salmonella, and Serratia marcescens and sequences from (ii) Photorhabdus luminescens, Yersinia, Erwinia, and Sodalis glossinidius, albeit the sequence of Sodalis glossinidius shows only a rudimentary similarity to those of the other three members of this group. Inverted repeats followed by a run of T’s, typical of Rho-independent terminators after the proBA operon, have been observed in E. coli (31) and Serratia marcescens (81). Except in Sodalis glossinidius, similar structures are present in the other organisms. In E. coli, Shigella, and Salmonella, the lengths of the inverted repeat and the runs of T's are shorter than those in Serratia marcescens, and it has not been shown experimentally whether these sequences function as terminators in any organism.

Fig. 4Transcription termination regions of the proline biosynthetic genes of enteric bacteria. The pairs of sequences highlighted with a yellow background are inverted repeats, with allowance for possible G-U pairing in the mRNA. Runs of T’s (U’s in mRNA) are highlighted in orange. Asterisks indicate nucleotides that are conserved across the species in each group. (A) Terminator regions of the proBA operon. The boxed TAA and TGA sequences in red are the translation termination codons of the proA gene. The sequences from E. coli and Shigella flexneri are identical in the region shown. Asterisks indicate nucleotides that are conserved in E. coli, Shigella flexneri, and S. enterica serovar Typhimurium (i) and in Photorhabdus luminescens, Yersinia pestis, and Erwinia carotovora (ii). Plus signs indicate nucleotides that are conserved between Erwinia carotovora and Sodalis glossinidius, and exclamation marks show nucleotides that are conserved among all four species listed in panel ii. (B) proC terminator regions. (i) For E. coli, Shigella flexneri, and S. enterica serovar Typhimurium, the boxed TGA sequence in red is the translation terminator for the proC gene and the boxed TCA sequence in red is the translation terminator for the downstream adrA (yaiC) gene, read toward proC. The sequences from E. coli and Shigella flexneri are identical in this region. (ii) For Sodalis glossinidius, Photorhabdus luminescens, Yersinia pestis, Erwinia carotovora, and Serratia marcescens (ii) and for Wigglesworthia glossinidia (iii), the boxed TGA and TGA sequences in red are the termination codons for proC and the boxed ATG sequence in green is the initiator for the next downstream gene. Asterisks indicate the nucleotides conserved among Sodalis glossinidius, Photorhabdus luminescens, Yersinia pestis, and Erwinia carotovora, plus signs indicate nucleotides conserved between Erwinia carotovora and Serratia marcescens, and exclamation marks indicate nucleotides that are conserved among all five of these species. nt, nucleotides.

Source: The inverted repeat sequences were identified with the program Stemmer.

Comparison of the regions downstream of proC (Fig. 4B) suggests that these may be classified into three groups based on their sequence relatedness: those from (i) E. coli, Shigella, and Salmonella; (ii) Sodalis glossinidius, Photorhabdus luminescens, Yersinia, Erwinia, and possibly Serratia marcescens (although the sequence from the latter shows very limited similarity to those from the other three organisms in this group); and (iii) Wigglesworthia glossinidia, which once again is an outlier. The sequences in E. coli, Shigella, and Salmonella contain an inverted repeat followed by a T-rich region that may constitute the terminator. There are T-rich regions downstream of proC in Erwinia, Yersinia, Sodalis glossinidius, Photorhabdus luminescens, Serratia marcescens, and Wigglesworthia glossinidia, but these are not preceded by clear-cut inverted repeats, and so it is not obvious what the proC terminators may be in these organisms. In view of the proximity of the next downstream gene (the yggT ortholog in Erwinia, Yersinia, Sodalis glossinidius, Photorhabdus luminescens, and Serratia marcescens and the yggV ortholog in Wigglesworthia glossinidia), it is possible that proC may be cotranscribed with the subsequent gene in these species.

REGULATION OF PROLINE SYNTHESIS

Control of Intracellular Proline Levels

It was observed that, in cells grown in complex media, the cytoplasmic pool sizes of proline and glutamate increase linearly with the external osmolality (17, 75). However, in cells grown in proline-free media, osmotic stress caused a ~10-fold increase in the cytoplasmic glutamate pool but the levels of proline were low and invariant (23, 24, 112). These results suggest that the increase in proline levels at high osmolality seen in complex media is the result of stimulation of uptake of this metabolite rather than its synthesis. The fact that the accumulation of proline was unaffected by osmotic stress even though there was a large increase in the glutamate pool may be the result of end product inhibition of GK. Inactivation of proline catabolism by a proline dehydrogenase (putA) mutation resulted in an approximately twofold increase in the proline level in S. enterica serovar Typhimurium grown in glucose-minimal medium, suggesting that there is simultaneous synthesis and degradation of proline (24). Although the level of the endogenously synthesized proline pool is not elevated in response to osmotic stress in the wild-type strains, the intracellular proline concentration increases with increasing external osmolality in proline-overproducing mutants (24). This apparent osmotic regulation of proline accumulation in the overproducing strains is probably not due to stimulation of synthesis, but rather, it may be the result of increased retention or recapture of proline, achieved by osmotic stimulation of the ProP or ProU proline transport systems.

The Proline Biosynthetic Genes Are Transcribed Constitutively

The levels of the proline biosynthetic enzymes and the expression of proB-, proA-, or proC-lacZ fusions were essentially invariant regardless of proline starvation or excess (14, 16, 21, 49, 51, 81, 98). Thus, proline is atypical among amino acids in that the expression of its biosynthetic enzymes is not subject to transcriptional control. Feedback inhibition of the first enzyme appears to be the major, if not the sole, control mechanism of the pathway.

Toxic Proline Analogs

Various proline analogs have been shown to interfere with proline metabolism (Table 3). Some of these analogs can be taken up by the proline transport systems and therefore may compete with the uptake of proline, some mimic proline as a feedback inhibitor of GK, and some can be charged to prolyl-tRNA and incorporated into proteins. A number of these compounds can be broken down by proline dehydrogenase, and therefore their efficacy as proline antagonists may be increased by putA mutations (33, 121). There is a procedure for the synthesis of [³H]-L-azetidine-2-carboxylate (116).

Table 3Proline analogs

Mutations Affecting Feedback Regulation of γ-Glutamyl Kinase

L-Azetidine-2-carboxylate, 3,4-D,L-dehydroproline, and thiazolidine-4-carboxylate have been used for the isolation of mutations that decrease the sensitivity of GK to feedback regulation (21, 23, 111, 115). The rationale behind this approach is that proline produced at high levels can outcompete the incorporation of the toxic analogs into proteins, and selection of proline analog resistance would be expected to yield proline-overproducing strains even with those compounds that are not inhibitors of GK. However, because the analogs are also substrates for various proline transport systems, loss-of-function mutations in permease genes are ~100-fold more frequent than mutations that result in proline overproduction (22). Proline-overproducing mutants can be recognized by the criteria that they excrete proline or show increased salinity stress tolerance (21, 40). Mutations giving rise to the most pronounced proline overproduction tend to be in the proB gene (21, 23), although mutations causing a weaker phenotype can be recovered in argD (see above) (13). The mutations D107N and E143A in the E. coli GK caused an increase of over two orders of magnitude in the apparent K_i of the enzyme for proline (26, 28, 80, 100, 107, 108). Similar mutations have been isolated in GKs from at least five other organisms (reviewed by Fujita et al. [40]). Although some of these mutations appear to be scattered throughout the enzyme, there is a preferential clustering of them in residues corresponding to amino acids 137 to 148 of the E. coli GK, suggesting that this region may constitute part of the proline binding site (40). The result that proline overproduction can be achieved by mutations that decrease the sensitivity of GK to proline inhibition provides strong evidence that feedback inhibition is important for the regulation of proline synthesis in vivo.

A mutant of Serratia marcescens with a high level of proline overproduction was obtained after four rounds of mutagenesis and selection for resistance to three proline analogs (111). This derivative proved to have two mutations: substitution of an alanine for valine at residue 117 (V117A) in GK, resulting in a 700-fold decrease in feedback inhibition of the enzyme, plus a single-base-pair change in the promoter (80) that caused a fourfold increase in expression of the proBA operon (see above). Although by itself the promoter mutation was not sufficient to confirm detectable proline excretion or analog resistance, it may have contributed to the appearance of the proline analog resistance at one of the steps of selection.

Proline overproduction is associated with some growth rate impairment in media of low osmolality, which is exacerbated if the feedback-insensitive GK is expressed from a high-copy-number plasmid (73). It is not clear whether this problem is due to a limitation in the supply of glutamate, ATP, or NADPH, an excessive drain on carbon, nitrogen, or energy, an accumulation of a toxic product, or a futile cycle arising from the synthesis and catabolism of proline.

Data on the use of a genetically engineered E. coli strain for the overproduction of L-proline are available. A yield of 27 g of proline per liter with 40% conversion of glucose has been reported (14). This productivity was achieved by using a putA (proline dehydrogenase) mutant strain that carried a feedback insensitive mutant version of the proB gene (E143A) and with wild-type copies of the proA and proC genes on a high-copy-number plasmid. Production of 56 g of proline/liter has been reported for a putA mutant of Serratia marcescens carrying the combination of the V117A feedback-insensitive proB mutation and the up-promoter mutation (80), plus an uncharacterized mutation that resulted in a fourfold increase in the glutamate dehydrogenase (111). The role of the latter mutation has not been determined, but it was suggested that it may have contributed to proline production by elevating the intracellular glutamate concentration (111).

HYDROXYPROLINE

E. coli and Salmonella normally do not synthesize hydroxyproline, but it has been possible to engineer derivatives that make trans-4-hydroxy-L-proline or cis-3-hydroxy-L-proline by the expression of the Dactylosporangium sp. strain RH1 proline-4-hydroxylase (106) or the Streptomyces sp. strain TH1 proline-3-hydroxylase (79), respectively. These two enzymes are 2-oxoglutarate-dependent dioxygenases that use proline as substrate and produce hydroxyproline, CO₂, and succinate. The 2-oxoglutarate required for this reaction can be supplied by the tricarboxylic acid cycle or by catabolism of exogenously provided proline. The production of trans-4-hydroxyproline was boosted ~20-fold by the introduction of the proB74 feedback-insensitive mutant form of GK and a putA mutation into the E. coli strain expressing the Dactylosporangium proline 4-hydroxylase (106).

PROLYL-tRNA SYNTHETASE

Prolyl-tRNA synthetase (ProRS), which is specified by the proS gene, carries out the last reaction prior to the incorporation of proline into proteins: the ATP-dependent charging of proline to tRNAs. A subunit molecular mass of 47,000 Da has been proposed for the E. coli enzyme on the basis of centrifugation in density gradients (67), but the molecular mass deduced from the DNA sequence of the proS gene is 63,701 Da (38). The enzyme can be found in vitro as an inactive monomer and as an active dimer that is stabilized by ATP or tRNA. Its K_m values are 0.03 mM for ATP and 0.23 mM for proline (85). The enzyme can esterify the proline analogs L-thiazolidine-4-carboxylate, allo-4-hydroxy-L-proline, and L-azetidine-2-carboxylate. Open-chain compounds having structural similarities with portions of the proline molecule, such as N-methyl-, N-ethyl-, and N-propylglycine and N-methyl-L-alanine, are also charged to tRNA^Pro (84). Typically, tRNA synthetases have two error correction functions: pretransfer editing, which involves hydrolysis of an incorrect amino acid from the amino acyl-AMP complex, and posttransfer editing, which consists of deacylation of tRNAs that have been charged with an incorrect amino acid. The E. coli ProRS can carry out the former editing reaction in the absence of tRNA, which is not observed with other tRNA sythetases (120). A 180-amino-acid domain, which is found in most prokaryotic ProRS enzymes, has been identified to be important for both editing functions of the enzyme (120). ProRS enzymes from a variety of organisms can also add cysteine to prolyl-tRNA in vitro, albeit at a rate much lower than that seen with proline (2). Unlike other amino acids, mischarged cysteine is not edited out efficiently from the prolyl-tRNA by the E. coli ProRS (2).

The proS gene was originally cloned by complementing a temperature-sensitive mutation in this gene (15). The amino acid sequence of the enzyme shows 45 and 37% homologies with threonyl-tRNA synthetase and seryl-tRNA synthetase sequences, respectively. This comparison was instrumental for the discovery of three conserved motifs that subsequently were also detected in some of the other aminoacyl-tRNA synthetases and thus led to the definition of class II synthetases (38). The substrate specificity of the E. coli enzyme was expanded by the C443G substitution in the E. coli ProRS so that it could recognize the bulky proline analog (2S)-piperidine-2-carboxylate as a substrate (62).

The proS gene is preceded by a σ⁷⁰ promoter and followed by a Rho-independent termination signal. ProRS belongs to the group of seven synthetases that are transiently derepressed under conditions of starvation for the cognate amino acid. In the case of proline, derepression is three- to fourfold (5). The molecular mechanisms underlying the derepression effect have been studied for some synthetases (43a) but not for prolyl-tRNA synthetase.

INCORPORATION OF PROLINE ANALOGS INTO PROTEINS

A number of proline analogs can be incorporated into proteins in vivo or in vitro (Table 3). Kim et al. (62) carried out an extensive study of proline analogs as substrates in protein synthesis in E. coli by determining their efficiency of incorporation into elastin, which is made up of 20% proline residues. They found that (2S,4S)-4-fluoroproline, (2S,4R)-4-fluoroproline, and (2S)-3,4-dehydroproline could be incorporated efficiently into elastase in an E. coli proC mutant under conditions of proline starvation, which was imposed in order to reduce competition from the natural imino acid. Although (2S,4S)-4-hydroxyproline, (2S,4R)-4-hydroxyproline, (2S)-4,4-difluoroproline, and L-azetidine-2-carboxylate could not be used for protein synthesis under the conditions described above, they were incorporated into elastase in a strain that expressed the wild-type ProRS at an elevated level. Increased osmolality, which enhances the uptake of proline, stimulated the utilization of each of the above-mentioned proline analogs. (4R)-1,3-Thiazolidine-4-carboxylate could be incorporated into elastase under these conditions only in a mutant in which proline dehydrogenase was blocked by a putA mutation. Finally, (2S)-piperidine-2-carboxylate (pipecolate) could support the synthesis of elastase at high osmolality but also required the coexpression of the mutant form of ProRS with a relaxed substrate specificity due to the C443G mutation.

PROLYL-tRNA

E. coli and (all serovars of) S. enterica have three prolyl-tRNAs: proK, which recognizes the codon CCG; proL, which reads CCC and CCU; and proM, which translates CCA, CCG, and CCU. The sequences of orthologous prolyl-tRNAs are 100% identical between the two species, and there is a high degree of interspecies conservation in the sequences of the promoters, terminators, and other transcriptional control sites (data not shown). However, a duplication of 108 nucleotides from the 3' end of the proK gene through the terminator, which was observed in E. coli (65), is not present in S. enterica serovar Typhimurium. Of the 77 nucleotides in the primary sequences of the three prolyl-tRNAs, 53 are identical (Fig. 5). Presumably, the nucleotides that are not conserved are not crucial for recognition by ProRS. Nucleotides that are important for recognition by the charging enzyme were analyzed in the proK and proM tRNAs by site-directed mutagenesis (44, 124).

Fig. 5Comparison of the three prolyl-tRNA genes of E. coli and S. enterica serovar Typhimurium. The sequences of the respective three genes are 100% identical between the two organisms. The sequences highlighted with various colored backgrounds are inverted repeats that can form stem structures. Asterisks indicate nucleotides that are conserved in all three tRNAs.

PROLINE TRANSPORT

There are three major proline transport systems in E. coli and Salmonella: PutP, ProP, and ProU (Table 1). PutP is needed for the transport of proline when it is a carbon or nitrogen source, and accordingly, it is induced by exogenous proline (95, 96). ProP and ProU also take up proline, but their major function appears to be to transport osmoprotectants such as glycine betaine (N,N,N-trimethylglycine) and other structurally related compounds (25), and accordingly, their activities are increased under high osmolality. PutP or ProP can support the uptake of proline as an auxotrophic supplement in all media (23, 43), but in wild-type strains ProU can do so only at high osmolality because it is not expressed sufficiently otherwise. However, high osmolality is not necessary for ProU activity because in strains that express it constitutively, this system can take up proline rapidly enough to support protein synthesis (35). Proline auxotrophic mutants that lack all three proline transport systems can grow normally with ≥0.2 mM proline, suggesting that there is at least one other unidentified transport system that can recognize it when it is present at high concentrations (23, 43). Proline auxotrophic mutants that lack the three major proline transport systems can also grow rapidly if proline is provided in the form of a peptide, such as glycyl-proline. Finally, there is one more proline transport system, ProY, which is cryptic in the wild-type strain. A regulatory mutation at a locus called proZ, which was isolated as a suppressor of the catabolic defect of putP mutants, resulted in high level expression of the ProY protein and enabled the strain to catabolize proline as nitrogen source (69). Although ProY, when expressed at a high level, can supplant PutP for proline uptake, the physiologically relevant substrate of this permease remains to be established.

PutP

PutP is a 54-kDa integral membrane protein that mediates the uptake of proline with Na⁺ cotransport. There has been considerable progress on the elucidation of the transport mechanism of PutP and on the identification of residues that are involved in the binding of Na⁺ or proline (56, 57, 58, 59, 88, 89, 91, 92, 93, 94, 125).

ProP

The ProP system was discovered as a minor proline permease in putP mutants (77, 109). The activity of this transport system was found to be stimulated approximately threefold by starvation for proline or other amino acids (3). The energy for transport via ProP is provided by H⁺ cotransport (72).

The most noteworthy feature of ProP is that its activity is stimulated by osmotic stress. In their studies of proline transport in membrane vesicles, Kaback and Deuel (60) noted that incubation of the vesicles in buffers of high osmolality greatly enhances the proline transport activity. Although at that time, it was believed that this was due to stimulation of the PutP activity, it was subsequently shown that the osmotic stimulation of proline transport in whole cells is due to activation of ProP (22, 36). Milner et al. (78) demonstrated that the activity of the ProP protein in membrane vesicles is enhanced by hyperosmotic shifts, providing evidence that the osmotic activation occurs through a posttranslational mechanism.

Extensive progress has been made on biochemical characterization of the osmotic activation of ProP (reviewed by Poolman et al. [90] and Wood et al. [122]). A C-terminal His₆-tagged version of the ProP protein was purified and was used to reconstitute a functional transporter in artificial lipid vesicles. The transport activity of the protein was stimulated in the vesicles by osmotic upshift imposed by impermeant solutes in a manner qualitatively similar to that seen in whole cells, demonstrating that the ProP protein itself is an osmosensory molecule. The regulatory stimulus for the increase in ProP activity has been proposed to be the combination of increased ion concentration and macromolecular crowding in the vesicles that result from a decrease in their volume upon hyperosmotic shifts (27). The ProP protein contains a long C-terminal tail, which can form a coiled-coil structure and may promote the dimerization in vivo. Site-directed mutagenesis indicated that the dimerization of the ProP protein via this C-terminal tail is important for the osmotic activation of transport activity. Although osmotic activation of ProP has been demonstrated in liposomes, this response requires at least one additional factor in vivo, because a Tn5 insertion has been isolated in a gene designated proQ that caused a fivefold decrease in the high osmolality-dependent stimulation of ProP activity in whole cells. The ProQ protein may interact with ProP to regulate its activity, but the exact role of the former protein needs to be established.

ProU

The ProU system, like ProP, is primarily a transport system for osmoprotectants, including proline. ProU is an ABC-type transporter, made up of three components: ProV, which is a membrane-associated ATPase; ProW, which is a membrane-bound permease; and ProX, which is the periplasmic binding protein. ProU was discovered as a proline transporter (22), but subsequently it was shown to have a higher affinity for glycine betaine (19, 20) and other similar osmoprotectants. The regulation of ProU is primarily effected by transcriptional control, which has been shown to entail ~600-fold induction of the proU (proVWX) operon by high osmolality (10). The transcriptional control of the proU operon is mediated by an unusual mechanism for a prokaryotic system, which involves a silencer that is located within the proV gene, 100 to 300 nucleotides downstream of the promoter (83).

A crystallographic structure of the E. coli ProX protein has been determined, which revealed the binding site for glycine betaine and proline betaine (102). Although physiological studies indicated that ProU can transport proline, ectoine, and a variety of other osmoprotectants, ProX has high affinity only for glycine betaine and proline betaine (K_m values of 1 and 5 μM, respectively) (102) and does not have a detectable binding affinity for proline or ectoine (11, 55). Genetic analysis, however, indicated that the ProX protein is required for the transport of these substrates in vivo, suggesting that the binding protein may be needed for the functioning of the transport system but not necessarily for capturing the substrates in the periplasm.

The proU operon in S. enterica serovar Typhi strains contains a deletion of an A-T base pair corresponding to position 432 of the proV gene of other sequenced S. enterica serovars, resulting in a frameshift and termination of the product at codon 187 (86). It is not known whether the altered proV gene product has any function or the downstream proW and proX genes are expressed in S. enterica serovar Typhi. However, interestingly, the transcriptional control region of the proU operon, including the promoter and the silencer, is conserved perfectly in this serovar compared to those in the other Salmonella serovars.