Inclusion of S-sepharose beads in the culture medium significantly improves recovery of secreted rBPI(21) from transfected CHO-K1 cells

rBPI(23), a recombinant N-terminal fragment of human bactericidal/permeability-increasing protein (BPI), kills gram-negative bacteria and binds endotoxin. rBPI(21), a variant, in which cysteine 132 is changed to alanine, retains the activities of rBPI(23). Initial attempts using conventional ion-exchange chromatography to purify rBPI(23) from culture supernatants of transfected CHO-K1 cells resulted in lower than expected yields. Also, ELISA of supernatants from CHO-K1 transfectants expressing rBPI(23) or rBPI(21) yielded variable signals. Results from pulse-chase experiments using [(35)S]methionine had indicated that rBPI(23) could not be detected in the culture medium by 7 h of chase, suggesting that these proteins were degraded and/or bound to cells, media components, or vessel surfaces. To address these issues, we developed a novel process whereby sterile S-Sepharose beads were added directly to the cell culture medium. For attached cells, the beads were added to confluent cultures with serum-free medium for the expression phase, while for suspension-adapted cells, beads were added at the beginning of culture growth. The S-Sepharose was then separated from cells and media and washed, and BPI was eluted with high-salt buffer. This approach yielded up to a 50-fold improvement in recovery of rBPI(23) and rBPI(21) from roller bottles, shake flasks, and 2-liter fermenters. It also resulted in improved detection and quantitation of secreted rBPI(23) and rBPI(21) by ELISA. Results of competition binding studies with iodinated rBPI(21) in conjunction with unlabeled rBPI(21) and rBPI(23) or with heparin demonstrated that these proteins bound specifically and with high affinity to heparan-containing sites on the surface of the CHO-K1 cells. We conclude that the S-Sepharose included in the culture medium captures the BPI protein products as they are secreted and protects them from degradation and/or irreversible binding to cell surfaces. This method has been scaled up to a manufacturing process in large (2750 liter) fermenters for pharmaceutical production.

Bactericidal/permeability-increasing protein inhibits growth of a strain of Acholeplasma laidlawii and L forms of the gram-positive bacteria Staphylococcus aureus and Streptococcus pyogenes

Bactericidal/permeability-increasing protein (BPI) inhibited growth of cell wall-deficient Acholeplasma laidlawii and L forms of certain strains of Staphylococcus aureus and Streptococcus pyogenes. However, the same strains of S. aureus and S. pyogenes with intact cell walls were not susceptible to the growth-inhibitory effects of BPI.

Biochemical characterization of recombinant fusions of lipopolysaccharide binding protein and bactericidal/permeability-increasing protein. Implications in biological activity

The physiological response to endotoxin (lipopolysaccharide (LPS)) can be regulated by two closely related LPS-binding proteins, LPS-binding protein (LBP), which potentiates LPS' inflammatory activity via interaction with the monocytic antigen CD14, and bactericidal/permeability-increasing protein (BPI), which neutralizes LPS. Both proteins bind LPS with high affinity sites in their N-terminal domains, whereas interaction between LBP and CD14 is dependent upon the LBP C-terminal domain. We have created fusions of the N- and C-terminal domains from each protein and compared the functional activities and pharmacokinetics of these fusions, the individual N-terminal domains, and the parent proteins. The N-terminal domains of BPI and LBP bound lipid A with their characteristic apparent affinity constants, regardless of the C-terminal fusion partner. In addition, the C-terminal domain of LBP allowed transfer of LPS to CD14 in conjunction with either N-terminal LPS binding domain. Proteins containing a BPI N-terminal domain had greater heparin binding capacities in vitro and were cleared more rapidly from the plasma of whole animals. Taken together, these data better define how closely related proteins such as BPI and LBP can have opposing effects on the body's response to LPS.

Expression and characterization of cysteine-modified variants of an amino-terminal fragment of bactericidal/permeability-increasing protein

rBPI23 is a biologically active, recombinant N-terminal fragment of human bactericidal/permeability-increasing protein (BPI). While rBPI23 is readily purified from culture supernatants of Chinese hamster ovary (CHO)-K1 transfectants, it is heterogeneous, consisting of monomer and disulfide-linked dimer, characteristics due presumably to the presence of three cysteines within the molecule. We have examined the role of these cysteines in rBPI23 expression, function, and dimer formation by mutating their codons to alanine (C132A), serine (C135S), or alanine (C175A) and expressing analogues of N-terminal fragments ("variants") lacking one, two, or all three cysteines in permanently transfected CHO-K1 cells. We also expressed a variant in which serine 18 was changed to cysteine (S18C), as found in both bovine and rabbit BPI. The C132A variant was readily secreted and purified as a homogeneous, stable monomeric protein species. The C135S and S18C variants were produced as mixtures of monomer and dimer; the C135S variant was poorly secreted, difficult to purify, and unstable on storage. In contrast, the C175A variant and those lacking any two or all three cysteines were expressed but not secreted. Purified rBPI23 and the C132A and S18C variants had comparable bactericidal and lipopolysaccharide (LPS) binding activities and were similarly effective at neutralizing LPS-induced tumor necrosis factor synthesis by THP-1 cells; the purified C135S variant lacked all activities. From these studies with CHO-K1 transfectants, we conclude that (i) cysteines 135 and 175 are both necessary for efficient secretion of a biologically active N-terminal BPI fragment, presumably through the formation of a disulfide bond, (ii) cysteine 132 is responsible for dimer formation, and (iii) only the C132A modification yields a stable, biologically active, N-terminal BPI fragment (designated rBPI21) that is free of dimeric species.

Human lipopolysaccharide-binding protein potentiates bactericidal activity of human bactericidal/permeability-increasing protein

Human bactericidal/permeability-increasing protein (BPI) from neutrophils and a recombinant amino-terminal fragment, rBPI23, bind to and are cytotoxic for gram-negative bacteria both in vitro and ex vivo in plasma or whole blood. To function in vivo as an extracellular bactericidal agent, rBPI23 must act in the presence of the lipopolysaccharide-binding protein (LBP), which also binds to but has no reported cytotoxicity for gram-negative bacteria. LBP, which is present at 5 to 10 micrograms/ml in healthy humans and at much higher levels in septic patients, mediates proinflammatory host responses to gram-negative infection. On the basis of these previous observations, we have examined the effect of recombinant LBP (rLBP) on the bactericidal activity of rBPI23 against Escherichia coli J5 in vitro. Physiological concentrations of rLBP (5 to 20 micrograms/ml) had little or no bactericidal activity but reduced by up to approximately 10,000-fold the concentration of BPI required for bactericidal or related activities in assays which measure (i) cell viability as CFUs on solid media or growth in broth culture and (ii) protein synthesis following treatment with BPI. LBP also potentiated BPI-mediated permeabilization of the E. coli outer membrane to actinomycin D by about 100-fold but had no permeabilizing activity of its own. Under optimal conditions for potentiation, fewer than 100 BPI molecules were required to kill a single E. coli J5 bacterium.

Chimeric immunoglobulin light chains are secreted at different levels: influence of framework-1 amino acids

Immunoglobulin light chain proteins are generally thought to be readily secreted without their corresponding heavy chains; non-secreted light chains have been viewed as aberrant forms. We have re-examined this assumption by expressing chimeric mouse-human light chains constructed for 12 mouse antibodies (mouse variable regions fused to a human kappa light chain constant region) in Sp2/0 and CHO cells. Five of the 12 light chains were either poorly secreted or not secreted at all. There was approximately a five-fold difference in the levels of secreted light chain between the highest poor secretor and the lowest good secretor. All of these light chains formed functional chimeric IgGs, which were secreted at similar levels, when co-expressed with their respective chimeric mouse-human heavy chains (mouse variable regions fused to a human gamma-1 heavy chain constant region). The influence of variable region amino acids on light chain secretion was examined by replacing the Framework-1 region of three poorly-secreted chimeric light chains with that of a readily-secreted light chain. For two of the light chains, secretion levels increased approximately 30- and 100-fold relative to that of the unmodified light chains. Comparison of the Framework-I amino acid sequence of the poorly- and readily-secreted light chains revealed an asparagine (N) and proline (P) at positions 11 and 12, respectively of these poorly-secreted light chains and a leucine (L) and serine (S) in the same region for some of the readily secreted light chains. Alteration of the NP to LS for one of the poorly-secreted light chains resulted in an approximately seven-fold increase in light chain secretion over that of the native form of the poorly-secreted light chain. We conclude from these studies that poor secretion can be a naturally occurring state for normal light chains and that amino acids within Framework-1 contribute to poor secretion for some of the light chains.

Human-engineered monoclonal antibodies retain full specific binding activity by preserving non-CDR complementarity-modulating residues

Humanization of murine monoclonal antibodies for human therapy has commonly been achieved by complementarity-determining region (CDR) grafting, in which murine CDR loops are grafted onto human framework regions. Difficulties with that method have revealed the importance of certain framework residues in determining both the 3-D structure of CDR loops and the overall affinity of the molecule for its specific ligand. In the general model of structure-function relationships presented here, each amino acid position in the variable region is classified according to the benefit of achieving a more human-like antibody versus the risk of decreasing or abolishing specific binding affinity. Substitutions of human residues at low-risk positions (exposed to solvent but not contributing to antigen binding or antibody structure) are likely to decrease immunogenicity with little or no effect on binding affinity. Changes at high-risk positions (directly involved in antigen binding, CDR stabilization or internal packing) are avoided to preserve the biological activity of the antibody. Moderate-risk changes are made with caution. This model has been tested experimentally using H65, an anti-CD5 murine monoclonal antibody, whose binding activity had been greatly reduced by two previous attempts at humanization by conventional CDR grafting. The new 'human-engineered' H65 antibody containing 20 low-risk human consensus substitutions (expressed as either IgG or Fab) retains the full binding avidity of parental murine and chimeric H65 antibodies. A human-engineered antibody with an additional 14 moderate-risk substitutions has unexpectedly enhanced avidity (3- to 7-fold). This method is generally applicable to the design of other human-engineered antibodies with therapeutic potential.

An amino-terminal fragment of human lipopolysaccharide-binding protein retains lipid A binding but not CD14-stimulatory activity

LPS-binding protein (LBP) mediates the pro-inflammatory effects of bacterial LPS by enhancing LPS-induced cytokine production by monocytic cells. LBP binds specifically to LPS to generate a complex that interacts with the CD14 receptor on the surface of responsive cells. To identify the biologically active regions of the protein responsible for mediating these activities, we cloned and expressed human rLBP (456 amino acids) as well as a truncated form encoding amino acids 1-197 (rLBP25). Both forms of LBP bound to LPS with the same affinity, and similarly inhibited LPS activity in the Limulus amebocyte lysate assay. These results demonstrate that the LPS-binding domain of LBP resides entirely within the N-terminal 197 amino acids of the protein. rLBP and rLBP25 were compared for their ability to mediate CD14-dependent LPS effects on cells. rLBP was effective in mediating uptake of LPS and stimulation of TNF production by human monocytic THP-1 cells, whereas rLBP25 had no significant activity in these assays. Similarly, rLBP was able to mediate LPS-induced TNF production by human PBMC whereas rLBP25 was essentially inactive. These results suggest that the structural features of LBP required for mediating LPS effects via CD14 are probably located in the C-terminal region of the protein. Thus, the LPS-binding activity of LBP can be separated from the CD14-stimulatory activity, suggesting these activities are mediated by structural elements residing in different regions of the protein.

Competition between rBPI23, a recombinant fragment of bactericidal/permeability-increasing protein, and lipopolysaccharide (LPS)-binding protein for binding to LPS and gram-negative bacteria

Lipopolysaccharide (LPS)-binding protein (LBP) and bactericidal/permeability-increasing protein (BPI) are two structurally related lipid A-binding proteins with divergent functional activities. LBP mediates activation of macrophage and other proinflammatory cells. In contrast, BPI has potent bactericidal and LPS-neutralizing activities. A recombinant fragment of BPI (rBPI23) retains the potent biological activities of the holo protein and may represent a novel therapeutic agent for the treatment of gram-negative infections, sepsis, and endotoxemia. For therapeutic effectiveness in many clinical situations, rBPI23 will have to successfully compete with high serum levels of LBP for binding to endotoxin and gram-negative bacteria. The relative binding affinities of rBPI23 and human recombinant LBP (rLBP) for lipid A and gram-negative bacteria were evaluated. The binding of both proteins to lipid A was specific and saturable with apparent Kds of 2.6 nM for rBPI23 and 58 nM for rLBP. rBPI23 was approximately 75-fold more potent than rLBP in inhibiting the binding of 125I-rLBP to lipid A. The binding affinity of rBPI23 (Kd = 70 nM) for Escherichia coli J5 bacteria was also significantly higher than that of rLBP (Kd = 1,050 nM). In addition, rBPI23 at 0.2 micrograms/ml was able to inhibit LPS-induced tumor necrosis factor release from monocytes in the presence of 20 micrograms of rLBP per ml. These results demonstrate that rBPI23 binds more avidly to endotoxin than does rLBP and that, even in the presence of a 100-fold weight excess of rLBP, rBPI23 effectively blocks the proinflammatory response of peripheral blood mononuclear cells to endotoxin.

An amino-terminal fragment of human lipopolysaccharide-binding protein retains lipid A binding but not CD14-stimulatory activity

LPS-binding protein (LBP) mediates the pro-inflammatory effects of bacterial LPS by enhancing LPS-induced cytokine production by monocytic cells. LBP binds specifically to LPS to generate a complex that interacts with the CD14 receptor on the surface of responsive cells. To identify the biologically active regions of the protein responsible for mediating these activities, we cloned and expressed human rLBP (456 amino acids) as well as a truncated form encoding amino acids 1-197 (rLBP25). Both forms of LBP bound to LPS with the same affinity, and similarly inhibited LPS activity in the Limulus amebocyte lysate assay. These results demonstrate that the LPS-binding domain of LBP resides entirely within the N-terminal 197 amino acids of the protein. rLBP and rLBP25 were compared for their ability to mediate CD14-dependent LPS effects on cells. rLBP was effective in mediating uptake of LPS and stimulation of TNF production by human monocytic THP-1 cells, whereas rLBP25 had no significant activity in these assays. Similarly, rLBP was able to mediate LPS-induced TNF production by human PBMC whereas rLBP25 was essentially inactive. These results suggest that the structural features of LBP required for mediating LPS effects via CD14 are probably located in the C-terminal region of the protein. Thus, the LPS-binding activity of LBP can be separated from the CD14-stimulatory activity, suggesting these activities are mediated by structural elements residing in different regions of the protein.

Potent anti-CD5 ricin A chain immunoconjugates from bacterially produced Fab' and F(ab')2

We have used genetic engineering to obtain secretion of anti-human CD5 antibody fragments from Escherichia coli for conjugation to the 30-kDa form of ricin A chain (RTA30). This was accomplished by introducing stop codons at two positions in the hinge region of the human IgG1 gene so that coexpression of the truncated heavy-chain genes (Fd') with a light chain would result in Fab' and/or F(ab')2 proteins containing either one or two interheavy-chain cysteines. An Fd' gene encoding both interheavy-chain cysteines yielded a mixture of F(ab')2 and Fab', which could be separated by size-exclusion chromatography. An Fd' gene encoding only one interheavy-chain cysteine yielded primarily Fab'. Purified F(ab')2 protein was equivalent to unlabeled chimeric IgG in competing for binding of IgG with CD5 antigen, while the molar concentration of the monovalent Fab' required for 50% binding inhibition was 4- to 5-fold higher than IgG. An immunoconjugate was prepared with Fab' by direct coupling to the unique free cysteine on RTA30. The bivalent F(ab')2 was conjugated to RTA30 after derivatization with the crosslinking agent 5-methyl-2-iminothiolane. These immunoconjugates efficiently killed a CD5+ T-cell line and human peripheral blood T cells.

Secretion of functional antibody and Fab fragment from yeast cells

We have constructed yeast strains that secrete functional mouse-human chimeric antibody and its Fab fragment into the culture medium. For chimeric whole antibody, cDNA copies of the chimeric light-chain and heavy-chain genes of an anti-tumor antibody were inserted into vectors containing the yeast phosphoglycerate kinase promoter, invertase signal sequence, and phosphoglycerate kinase polyadenylylation signal. Simultaneous expression of these genes in yeast resulted in secretion of properly folded and assembled chimeric antibody that bound to target cancer cells. Yeast chimeric antibody exhibited antibody-dependent cellular cytotoxicity activity but not complement-dependent cytotoxicity activity. For production of Fab fragments, a truncated heavy-chain (Fd) gene was created by introducing a stop codon near the codon for the amino acid at which papain digestion occurs. Simultaneous expression of the resulting chimeric Fd and light-chain genes in yeast resulted in secretion of properly folded and assembled Fab fragment that bound to target cancer cells.

Escherichia coli secretion of an active chimeric antibody fragment

A chimeric mouse-human Fab protein that binds specifically to the human carcinoma cell line C3347 has been expressed and secreted from Escherichia coli. This molecule, which contains functionally assembled kappa and Fd proteins, binds as effectively to sites on the surface of C3347 cells as Fab fragments prepared proteolytically from whole chimeric or mouse antibody. The production in Escherichia coli of foreign heterodimeric protein reagents, such as Fab, should prove useful in the management of human disease.

Functional limits of the araIc promoter suggest an additional regulatory site for araBAD expression

The araBAD promoter is defined, in part, by two types of cis-acting constitutive mutations, araIc at position -35 and araXc at position -10. Subcloning experiments demonstrated that the araIc and araIcXc promoters require DNA sequence information out to position -53 to -56 for maximum constitutive expression. This is 8 to 10 base pairs more DNA than is generally thought to be necessary for RNA polymerase interaction. The -53 to -56 region is required for glucose repression, suggesting that an additional factor interacts in this region and is necessary for maximum expression.

DNA sequence of the araBAD-araC controlling region in Salmonella typhimurium LT2

The araB and araC genes of Salmonella typhimurium have been cloned onto the plasmid pBR322. Restriction analysis and subcloning of restriction fragments localized these genes to a 4.4 kb DNA fragment. Complementation analysis revealed that the cloned araB and araC genes from S. typhimurium complemented araB and araC mutant strains of escherichia coli. Conversely, cloned araB and araC genes from E. coli complemented araB and araC mutant strains of Escherichia coli. Conversely, cloned araB and araC genes from E. coli complemented araB and ara C mutant strains of S. typhimurium. The DNA sequence was determined for the S. typhimurium araB and araC controlling region and for the initially translated portions of these genes. The nucleotide sequence of the araB promoter was 87% homologous with the same region in E. coli and contained no deletions or insertions relative to the E. coli sequence. The presumed AUG codon corresponding to the amino terminus of the S. typhimurium araC protein was in the same location as in E. coli. There was, however, considerable divergence for the E. coli sequence preceding the translation start site. The nucleotide sequence of the initial 237 bp in the open reading frame of the S. typhimurium araC gene was 78% homologous with the same sequence in E. coli. By comparison, the amino acid sequence for this region was 91% conserved.

The araIc mutation in Escherichia coli B/r

The araIc allele is a cis-acting mutation which has been used to define the araBAD promoter in Escherichia coli B/r. Nineteen araIc mutants were originally isolated by Englesberg and co-workers as Ara+ "revertants" of an araC deletion mutant (Englesberg et al. J. Mol. Biol. 43:281-298, 1969). The mutants constitutively expressed araBAD gene products in the absence of functional araC activator protein. Eight of the araIc mutations have been cloned by in vivo recombination onto pBR322-ara hybrid plasmids. Restriction and DNA sequence analysis of these araIc mutations showed that they result from a single base-pair change located at -35 in the araBAD promoter.

Characterization of a site-specific restriction endonuclease from Streptomyces aureofaciens

A new type II sequence-specific restriction endonuclease, SauI, was isolated from Streptomyces aureofaciens IKA18/4. The purified enzyme was free of contaminating exonuclease and phosphatase activities. SauI cleaved lambda DNA at two sites, but did not cleave pBR322, simian virus 40, or phi X174 DNA. SauI recognized the septanucleotide sequence 5'-CCTNAGG-3' and cleaved at the position indicated by the arrow, producing a trinucleotide 5'-terminal extension.

DNA sequence of the araC regulatory gene from Escherichia coli B/r

The DNA sequence of the araC regulatory gene from Escherichia coli B/r has been determined by the base-specific chemical cleavage reactions of Maxam and Gilbert. An open reading frame is found which codes for a protein of 292 amino acids. A nonsense mutation, araC5, is shown to result from a G to A transition at nucleotide 429 converting the tryptophan codon TGG to the amber codon TAG. A deletion which does not recombine with any known point mutation in araC, delta(araCO)719, removes all but the last 22 codons of the gene.

The araC regulatory gene mRNA contains a leader sequence

An estimation of the size of the araC gene in Escherichila coli B/r was made by sub-cloning restriction fragments of the araC-containing hybrid plasmid pTB1 into the plasmid pBR322. Plasmids which contained a functional araC gene were identified by genetic complementation tests. DNA sequence analysis of the promoter-proximal region of the araC gene revealed that araC mRNA contains a 150 nucleotide leader.

Construction of pBR322-ara hybrid plasmids by in vivo recombination

In vivo recombination was used to clone deletions of the araBAD-araC genes of Escherichia coli onto a hybrid pBR322-ara plasmid. Genetic and physical analyses demonstrated that the desired deletions had been recombined onto the plasmid. In addition to permitting a detailed physical analysis of various ara deletions, this procedure has generated a series of plasmid cloning vehicles that can be used to clone, by in vivo recombination, any ara point mutation located within the region covered by the deletions. Hybrid plasmids containing the cloned point mutation can be distinguished from the original cloning vehicle by genetic complementation. The desired recombinant plasmid can be easily obtained because the frequency of recombination between the plasmid ara region and the chromosomal ara region is 0.025%--3%. A plasmid containing a deletion which removes the ara controlling site region and the araC gene was used to clone two types of araBAD promoter mutations and an araC mutation by in vivo recombination. Genetic and physical analysis of these plasmids established that the mutations in question had been recombined on to the ara deletion plasmid. The application of this procedure to the ara genes and to other genetic systems is discussed.

Deoxyribonucleic acid sequence of araBAD promoter mutants of Escherichia coli

The controlling site region for the araBAD operon is defined, in part, by two classes of cis-acting constitutive mutations. The aralc mutations allow low-level constitutive expression of ara-BAD in the absence of the positive regulatory protein coded for by the araC gene, whereas the araXc mutations allow expression of araBAD in the absence of the cyclic adenosine monophosphate receptor protein. Six independently isolated aralc mutations and three independently isolated araXc mutations were cloned onto the plasmid pBR322 using in vitro recombinant deoxyribonucleic acid techniques and in vivo recombination between plasmid and chromosomal deoxyribonucleic acid. The location of these mutations was determined by deoxyribonucleic acid sequence analysis. All of the aralc mutations occurred at position -35 within the araBAD promoter (+1 = messenger ribonucleic acid start for araBAD) and resulted from an AT leads to GC transition. All of the araXc mutations occurred at position -10 within the araBAD promoter and resulted from a GC leads to AT transition. Models are presented to explain the mode of action of the aralc and araXc mutations.

Effects of magnesium, calcium, and serum on reversion of stable L-forms

The L-form of Agromyces ramosus was stable in the absence of penicillin when transferred on heart infusion agar containing NaCl and serum. It reverted to its bacterial form, however, when magnesium replaced the serum in this medium. On a dilute medium containing NaCl but lacking serum, the L-form died out unless calcium, magnesium, or serum was added. It grew as the L-form in the presence of calcium of serum but reverted to the bacterial form in the presence of magnesium. Reversion also occurred when magnesium was added to the dilute medium containing serum. Calcium interfered with or prevented the magnesium-induced reversion. The revertant bacterial form resulting from these studies was not NaCl sensitive, as was the case of the bacterial revertant of this organism produced in soil (A. H. Horwitz and L. E. Casida, Jr., Can. J. Microbiol, 24:50--55, 1978).

Survival and reversion of a stable L form in soil

The stable L form of Agromyces ramosus reverted to a bacterial form when incubated in sterilized soil. The cellular and colonial morphology of this bacterial form resembled that of the original parent bacterial form. The two forms differed, however, in that the revertant maintained its bacterial form when transferred onto a low-salt (NaCl) medium but was virtually completely induced into the L-form state on a high-salt medium. The original parent bacterial form was not sensitive to salt. The possibility is discussed that an L-form interchanging with a bacterial-form cycle for this bacterium might occur naturally in soil. This cycle would be mediated by fluctuations in local salt concentrations in the soil.

L-Phase variants of Agromyces ramosus

Agromyces ramosus, which is a numerically prevalent bacterium in soil, was easily induced into the L-phase by growing it on agar media containing low levels of penicillin or glycine. The L-forms were stable after initial contact with the inducing agent and could not be reverted to the bacterial form by any of the procedures tried. These results are discussed in relation to a possible natural occurrence of L-forms of this bacterium in soil.