INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

The key to success for many bacteria in causing infection is colonization of host tissues. Enteric bacteria, such as Salmonella spp., have to survive in harsh host microenvironments including the intestinal mucosa. At the intestinal mucosa, these bacteria encounter host defense mechanisms including antimicrobial peptides (AP), which are cationic, amphipathic molecules that kill bacteria by membrane permeabilization. Within the intestine, AP are secreted into the lumen by Paneth cells located in the base of intestinal crypts. AP are also found within phagocytic cells located in the intestinal submucosa. Following oral ingestion, typhoid fever-causing strains of Salmonella can transcytose through M cells and intestinal epithelial cells and are then taken up by and survive within resident phagocytes. The ability of salmonellae to survive within the host intestine and within professional phagocytes is likely to depend, at least in part, on mechanisms of resistance to AP.

Lipopolysaccharade (LPS) is the major surface component of gram-negative bacteria. Lipid A is the bioactive component of LPS that comprises the outer leaflet of the gram-negative bacterial outer membrane. Phosphate groups on lipid A and LPS core components result in the bacterial surface having a net negative charge. This charge plays a significant role in the electrostatic interaction of cationic AP with the bacterial surface. Gram-negative organisms can synthesize LPS with specific lipid A and core modifications in response to environmental conditions that include those found in host tissues. Lipid A modifications are best characterized for Salmonella spp. in which the addition of palmitate or 4-aminoarabinose (Ara4N) to lipid A is coordinately regulated by the two-component system PhoP-PhoQ (9-11), which is necessary for bacterial survival within macrophages and within the host (6, 12, 15). Ara4N-containing lipid A results in a less negatively charged bacterial surface, which reduces AP binding and promotes resistance to the cationic AP polymyxin B (PM), bactericidal/permeability-increasing protein, and azurocidin (17, 18).

PhoP-PhoQ-mediated Ara4N addition to lipid A occurs through transcriptional activation of the genes encoding another two-component system, PmrA-PmrB, which can itself be activated independent of PhoP-PhoQ by low-pH or high-iron conditions (9, 14, 20). Two loci, pmrE and pmrHFIJKLM, are regulated by PmrA-PmrB and are essential for both biosynthesis of lipid A containing Ara4N and for resistance to PM (1, 8). This study was conducted to examine the individual and collective roles of the genes of the pmrHFIJKLM operon, as well as flanking genes, in AP resistance, modification of lipid A, and Salmonella enterica serovar Typhimurium virulence.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Bacterial strains, culture conditions, and reagents. Bacterial strains used in this study are listed in Table 1. Cultures were grown on Luria-Bertani (LB) agar plates or broth at 37°C with aeration. Antibiotics were used at the following concentrations: chloramphenicol, 25 µg/ml; ampicillin, 50 µg/ml; kanamycin, 45 µg/ml; tetracycline, 15 µl/ml; streptomycin, 1,000 µg/ml. The chromogenic substrate for -galactosidase, 5-bromo-4-chloro-indolyl--D-galactopyranoside (X-Gal), was used at a concentration of 40 µg/ml.

Construction of luciferase fusions and transcription assays. pmrD was amplified from plasmid pKK01 (8) by PCR using primers JG177 (5' GGG AAT TCT GCC ATG TTC TGG TGC TGT GC 3') and JG178 (5' GGG GTA CCG CGC GTC AAC CGC TGC CAT TC 3'). This fragment was digested with the restriction enzymes EcoRI and KpnI and ligated into pGPL01, which is a Pir-dependent suicide vector containing a promoterless firefly luciferase gene downstream of the multiple cloning site (9). This clone, transformed into Escherichia coli SM10Pir, was mated with CS019, and a single recombinant was identified. P22HTint phage transduction was used to transduce the pmrD::luc fusion into various strains. Lucifierase activity was determined as previously described (9) after growth of strains to log phase (optical density at 600 nm [OD₆₀₀] ~0.6). A pmrM luciferase fusion was constructed in a manner identical to that for the pmrD fusion with primers JG252 (5' GGAATTCGGACTGATAAGCGTTGCG 3') and JG253 (5' GGGGTACCTGATGCACGCTGTTATTCC 3'). The luciferase fusion to the pmrHFIJKLM promoter was also constructed similarly to the pmrD and pmrM fusions, except that the promoter fragment was cloned into pLB02 (7), which contains DNA homologous to a region downstream of pagC, where the vector is recombined. The primers used to amplify the pmrHFIJKLM promoter were JG122 (5' GGGGTACCTGAAAGCCGCTTT TC 3') and JG123 (5' GGAATTCTTTTTACTTCACCT 3').

Analysis of LPS modification. LPS was isolated by Mg²⁺-ethanol precipitation as described by Darveau and Hancock (4), and lipid A was isolated by hydrolysis in 1% sodium dodecyl sulfate at pH 4.5 (3). Negative-ion matrix-assisted laser desorption ionization-time of flight (MALDI-TOF) mass spectrometry was performed as described previously (5). Lyophilized lipid A was dissolved with 5 µl of a 5-chloro-2-mercaptobenzothiazole (CMBT) MALDI matrix in chloroform-methanol (1:1 [vol/vol]) and then applied (1 µl) to the sample plate. All MALDI-TOF experiments were performed using a BiflexIII mass spectrometer (Bruker Daltonics, Inc., Billerica, Mass.). Mass spectroscopy tracings were analyzed for the presence or absence of Ara4N based on known peak locations at m/z 1,929 (hexa-acylated lipid A with Ara4N) and 2,167 (hepta-acylated lipid A with Ara4N).

Assays of DNA binding. PCR fragments of the pmrHFIJKLM and pmrCAB operons were digested with EcoRI (located at one end of each fragment) and labeled with [-³²P]dATP (10 µCi/reaction) with the Klenow fragment of DNA polymerase I. Binding reactions were carried out by the incubation of the purified PmrA or PmrA-constitutive (PmrA^c) His-tagged proteins with binding buffer (10mM Tris-HCl [pH 7.5], 1mM EDTA, 100 mM KCl, 0.1 mM dithiothreitol, 5% glycerol, 75 µg of sonicated salmon sperm DNA/ml, 10 mg of bovine serum albumin/ml for 5 min at room temperature. Labeled probe (1 to 5 ng) was added, and the reaction mixture was incubated for 10 min at room temperature. Samples were electophoresed at 4°C on a 5% acrylamide gel at 200 V. Gels were subsequently dried and autoradiographed.

Animal studies. Survival assays were accomplished as follows. Female BALB/c mice were inoculated orally with 20 µl of an overnight culture (washed and diluted to 10⁶ organisms/20 µl in phosphate-buffered saline (PBS), which is 1 log unit above the 50% lethal dose [LD₅₀]) using the animal's swallowing reflex. Mice were prefed 20 µl of 10% sodium bicarbonate 30 min prior to bacterial inoculation. Alternatively, after dilution in PBS, animals were inoculated via the intraperitoneal route with various numbers of organisms in a 100-µl volume. Overnight cultures were plated to enumerate the numbers of organisms inoculated. The average days of survival and the numbers of surviving mice were recorded. In competition assays, female BALB/c mice were inoculated orally as described above. Mice were inoculated with a mixture of 10 µl of two different strains (each containing 10⁶ organisms/10 µl). Mice were sacrificed when clearly moribund, and livers and spleens were removed. The organs were homogenized in 1 × PBS, diluted, and plated onto the appropriate antibiotic-containing agar plates, which selected for one of the two competing strains.

MIC assays of AP resistance. PM (U.S. Biochemicals; 8,040 U mg¹) was used at concentrations of 0.0084 to 12 µg/ml in both plate and broth assays. Standard MIC testing of susceptibility to PM was accomplished as described by Steinberg et al. (21).

Deletion analysis of the pmrHFIJKLM operon. Flanking regions of the genes of the operon were amplified using PCR and digested with the restriction enzyme NcoI. Primers that were used are listed in Table 2. The flanking regions were then ligated together and digested with the restriction enzymes EcoRI (3') and XbaI (5'). This fragment was then ligated into plasmid pKAS32, a suicide vector containing a dominant streptomycin sensitivity allele (19). Deletion constructs in SM10Pir were mated with JSG844, and single recombinants were selected by plasmid-encoded resistance to ampicillin and chromosome-encoded resistance to chloramphenicol and sensitivity to streptomycin. Losses of the plasmid (secondary recombination events) were selected following overnight incubation by growth on 1,000-µg/ml streptomycin-containing agar. Colonies were further screened by PCR to confirm the incorporation of the deletion. Mutation of the pmrM gene was accomplished not by the construction of a nonpolar deletion but by the construction of the pmrM::luc fusion described above, which essentially inserts plasmid pGPL01 into the pmrM coding sequence.

Confirmation of in-frame deletions and complementation experiments. PCR fragments overlapping the deletion sites were sequenced to confirm correct incorporation of the in-frame deletion. Deletions in pmrH and pmrF were complemented with plasmid pKK013 (carrying only the intact pmrH and pmrF genes). The pmrJ, pmrK, pmrL, and pmrM genes were complemented by pKK012-1 (carrying only the intact pmrJKL and pmrM genes). To further confirm nonpolarity, the pmrM::luc fusion was mated into each deletion strain and luciferase activity was monitored.

Isolation of the PmrA protein. The pmrA gene or the mutant, constitutive allele of this gene (pmrA505) was amplified by PCR with primers JG106 (5' GCGGATCCAAGATACTGATTGTTGAAG 3') and JG107 (5' AACTGCAGTTAGCTTTCCTCAGTGGC 3') from strains ATCC 14028s (wild type [WT]) and JSG435, respectively. These fragments were cloned into the N-terminal His tag-encoding vector pQE30 and transformed into strain XL1-Blue-MRF'. Transcription of the genes was induced with IPTG (isopropyl--D-thiogalactopyranoside) upon movement of the plasmids into strain M15/pREP4, and the His-tagged proteins were purified as previously described (7).

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

Definition of the pmrHFIJKLM transcriptional unit and promoter. The pmrHFIJKLM genes are unidirectionally transcribed, and the individual open reading frames are separated by no more than 5 bp, as shown in the diagram of this chromosomal region (Fig. 1). This arrangement suggested that these genes formed a typical prokaryotic operon. To demonstrate that these genes were indeed cotranscribed, a luciferase fusion to pmrM, the last gene of the operon, was constructed in single copy on the chromosome. Upon the movement of this fusion into PmrA-null (PmrA), PmrA^c, and PmrA^c pmrF::Tn10d backgrounds by P22-mediated transduction, luciferase activity was monitored. As shown in Fig. 2, this fusion was highly regulated by PmrA-PmrB, and the reporter activity in a PmrA strain was similar to that in a strain containing the pmrF::Tn10d insertion, demonstrating that the insertion in pmrF had a polar effect on downstream gene transcription. Therefore, these data strongly suggest that the pmrHFIJKLM genes are cotranscribed and that the promoter is upstream of this operon.

View larger version (9K): [in this window] [in a new window]   FIG. 1.   Diagram of the chromosomal region containing the pmrG, pmrD, and pmrHFIJKLM genes. Arrows, direction of transcription.

View larger version (29K): [in this window] [in a new window]   FIG. 2.   The pmrHFIJKLM genes are expressed as an operon. PmrA, PmrAc, and PmrAc pmrF::Tn10d strains were examined. Each strain contained a single-copy pmrM::luc fusion and was assayed in logarithmic phase (OD600, 0.6) for firefly luciferase-activity (expressed as relative light units [RLU]). The data are from a single experiment of three independent assays that gave similar results.

View larger version (44K): [in this window] [in a new window]   FIG. 3.   The pmrHFIJKLM promoter is highly PmrA-PmrB activated and tightly regulated. PmrA, PmrAc, PhoPc PmrA, and PhoP PmrAc strains containing a single-copy pmrHFIJKLM promoter::luc fusion were assayed in logarithmic phase (OD600, 0.6) for firefly luciferase activity (expressed as relative light units [RLU]). Note that the y-axis scale is logarithmic. The data are from a single experiment of three independent assays that gave similar results.

View larger version (47K): [in this window] [in a new window]   FIG. 4.   Binding of the PmrA-His and PmrAc-His proteins to the pmrHFIJKLM promoter (called pmrF) and the pmrCAB promoter (called pmrC). The sizes of the fragments used and the primers used to amplify the fragments are noted. Arrows, shifted fragments. N, no protein added; A, PmrA-His added; Ac, PmrAc-His added.

View larger version (50K): [in this window] [in a new window]   FIG. 5.   PmrA-His protein binding to the pmrCAB promoter is specific. The presence or absence of the PmrA-His protein, competitive DNA, and noncompetitive DNA is noted at the top. The numbers at the bottom are the lane numbers.

Regulation and role in PM resistance of the flanking genes of the operon. It has been shown that the gene upstream of and divergently transcribed from the pmrHFIJKLM operon is the PmrA-PmrB-regulated gene, pmrG (8). The gene directly downstream and divergently transcribed from the pmrHFIJKLM operon is pmrD, which imparts increased PM resistance to S. enterica serovar Typhimurium when expressed from a multicopy plasmid (16). To further the studies of this chromosomal region, experiments were designed to determine the potential regulation of pmrD and the role of pmrD and pmrG in resistance to PM.

View larger version (33K): [in this window] [in a new window]   FIG. 6.   PhoP-PhoQ but not PmrA-PmrB regulates the pmrD gene. PmrA, PmrAc, PhoPc, and Phop strains containing a single-copy pmrD::luc fusion were assayed in logarithmic phase (OD600, 0.6) for firefly luciferase activity (expressed as relative light units [RLU]). Results are expressed as ratios of firefly luciferase activity in PmrAc versus PmrA backgrounds and in PhoPc versus PhoP backgrounds.

Deletion analysis of the pmrHFIJKLM operon: roles of individual loci in LPS modification and PM resistance. To examine the roles of individual genes of the pmrHFIJKLM operon in resistance to PM and in Ara4N modification of lipid A, nonpolar deletions were constructed in pmrH, pmrF, pmrI, pmrJ, pmrK, and pmrL. An insertion mutation was created in pmrM by recombination of a suicide plasmid within this locus. The chromosomally incorporated deletions were confirmed to be nonpolar by sequencing the deletion junctions and by demonstrating the lack of an effect on luciferase activity of a pmrM::luc fusion mated into each deletion background. These nonpolar deletion or mutant strains (constructed in a PmrA^c background) were then examined for resistance to PM. All mutations except that in pmrM had a major effect on resistance to PM (MIC of PmrA^c and pmrM strains, 4 µg/ml; MIC of strains with pmrHFIJKL deletions, 0.1 µg/ml). To determine if the loss in resistance to PM was due to the loss of Ara4N in lipid A, MALDI-TOF mass spectroscopy was performed on purified lipid A. This analysis confirmed the absence of Ara4N in deletion strains with reduced resistance to PM (Fig. 7). This can be seen by the absence of peaks at m/z 1,929 and 2,167, which correspond to hexa- and hepta-acyl lipid A with Ara4N, respectively. Loss of the peaks at m/z 1,949 and 2,189 is also evident in the mutants, as these peaks correspond to hexa- and hepta-acyl lipid A, respectively, with Ara4N and the PhoP-PhoQ-mediated hydroxyl addition to myristate. The profile of the pmrM mutant was identical to that of the PmrA^c strain (the parental strain of all mutants examined [8]), further demonstrating that this gene plays no role in Ara4N addition to lipid A or PM resistance.

View larger version (18K): [in this window] [in a new window]   FIG. 7.   Mass spectroscopy of the lipid A from pmrF, pmrI, pmrK, pmrL, and pmrM nonpolar deletion strains. The m/z ratios and related structures are as follows: 1,797, unmodified hexa-acyl lipid A; 1,815, hexa-acyl lipid A with 2-OH myristate; 1,929, hexa-acyl lipid A with Ara4N; 1,949, hexa-acyl lipid A with 2-OH myristate and Ara4N; 2,036, unmodified hepta-acyl lipid A; 2,052, hepta-acyl lipid A with 2-OH myristate; 2,167, hepta-acyl lipid A with Ara4N; 2,189, hepta-acyl lipid A with 2-OH myristate and Ara4N. The lipid A from strains with pmrH and pmrJ nonpolar deletions was also examined, and the results were identical to those shown in panels A through D but were not included due to differences in the scale of m/z values examined and the difficulty in comparing these tracings to those for the other samples. The PmrAc strain that is the parental background of the strains with the examined deletions has been examined previously (8), and results for it are identical to those shown in panel E.

LPS modification affects oral virulence. It has been shown previously that a strain containing a pmrF::Tn10d insertion was unable to add Ara4N to lipid A, which resulted in a significant decrease in resistance to PM (8). To determine the effect on S. enterica serovar Typhimurium virulence of the lipid A modification with Ara4N, the pmrF::Tn10d mutant was examined in the mouse model of typhoid fever.

View larger version (20K): [in this window] [in a new window]   FIG. 8.   Mutations in pmrA or the pmrHFIJKLM operon affect virulence by the oral route. Mice were inoculated orally with a dose of 106 CFU. BALB/c mice were infected with ATCC 14028s (WT), WT with a pmrA::Tn10d insertion (PmrA), WT with a pmrF::Tn10d insertion (PmrF), or a combination of two of these strains. (A) Percent survival of BALB/c mice receiving the WT, PmrA, or PmrF strain. The numbers above the bars represent the number of surviving mice per the number of mice tested. (B) Percent survival of mice over 40 days postinfection. (C) Competition assays of WT versus PmrA and WT versus PmrF strains. The competitive index is the ratio of CFU in the livers of mice infected with a PmrA, WT, or PmrF strain to CFU in livers of mice infected with the WT strain. The WT-versus-WT experiment involved two WT strains marked with different antibiotic resistances. The competitions labeled in vitro were performed by growth overnight in LB medium to ensure that any in vivo effects were not due to a growth disadvantage. Both liver and spleen were examined in the competition assays, but only the data for the liver are shown.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

Activation of the PmrA-PmrB two-component regulatory system by environmental signals or by mutations within pmrA that result in a constitutive phenotype activates the transcription of genes whose products covalently modify LPS. These modifications (Ara4N and phosphoethanolamine) result in increased resistance to AP, which are membrane-active pore formers present at mucosal surfaces and within macrophages and neutrophils. Increased resistance to these peptides provides the bacterium with a survival advantage within the host. In this work, we presented a detailed characterization of an island of PmrA-PmrB- and PhoP-PhoQ-regulated genes. Experiments were conducted to examine the roles of the individual genes in this region in resistance to AP and what effect the loss of genes in this region has on S. enterica serovar Typhimurium virulence.

The pmrHFIJKLM locus is regulated by PmrA-PmrB, and a pmrF::Tn10d insertion was shown to eliminate the ability to modify lipid A with Ara4N and markedly reduced resistance to PM (8). These genes were shown to be cotranscribed, as a pmrM::luc fusion was activated by PmrA-PmrB and luciferase activity was dramatically reduced when the reporter fusion was recombined downstream of the pmrF::Tn10d insertion. Fusion of the promoter of this operon to the gene encoding firefly luciferase showed it to be over 3,000-fold activated by PmrA-PmrB. It is intriguing to propose that this highly expressed in vivo-activated promoter may be useful in a live-attenuated Salmonella vaccine for the expression of heterologous antigens. This approach has several advantages including high-level expression in host antigen-presenting cells and the ability to construct stable, single-copy chromosomal constructs. In fact, this approach has been successfully tested using phoP-regulated gene promoters (2, 13).

Studies of the regulators of two-component systems have shown that they bind to consensus sequences in the promoters of regulated genes. Therefore, experiments were designed to examine the interaction of PmrA with pmr promoters. The PmrA protein was purified (both WT and constitutive forms) and used in promoter binding studies. A 118-bp region of the pmrHFIJKLM promoter and a 40-bp region of the pmrCAB operon promoter were found to shift with each protein. The 40-bp region of the pmrCAB promoter is located just upstream of the 35 region, which places the putative binding site in a position comparable to those of other transcriptional activators of the OmpR family. Recently, Wøsten and Groisman (22) demonstrated similar binding of the PmrA protein to the promoters of activated genes and defined a consensus binding site of 5' TTAAKTTCTTAAKGTT 3' for PmrA. This site is located within both the pmrCAB and pmrHFIJKLM operon promoter fragments that bound PmrA (16 of 16 match in each). The PmrA^c protein bound approximately threefold more promoter fragments than the WT PmrA protein, suggesting that the constitutive activity of this protein may be due to increased promoter affinity.

Nonpolar deletions were constructed in each gene of the pmrHFIJKLM operon (except pmrM, which was disrupted by insertion), and the effect of these mutations on PM resistance was measured. All of the genes except pmrM had an effect on PM resistance, which resulted in an MIC level similar to that of a PmrA strain (MIC of PmrA^c and pmrM strains, 4 µg/ml; MIC of pmrHFIJKL deletion strains, 0.1 µg/ml). Based on protein homologies and known sugar biosynthesis pathways, a model involving the pmrHFIJKLM operon and pmrE has been proposed for Ara4N biosynthesis and addition to lipid A (1). This pathway begins with the conversion of UDP-glucose to UDP-glucuronic acid by the pmrE gene product, followed by the conversion to UDP-4-amino-L-deoxyarabinose and attachment of this moiety to lipid A by the products of the pmrHFIJKLM operon. This model, however, was unable to identify functions for PmrL and PmrM, as neither of these small proteins has any strong similarities to proteins in GenBank. From this study, it is clear that PmrL does belong at some point in the pathway but PmrM does not. Therefore, the function of PmrM remains unknown.

Previous studies showed that the gene upstream of the pmrHFIJKLM operon, pmrG, was regulated by PmrA-PmrB (8). Because the effect of this gene on the PM resistance phenotype had not been previously examined, a pmrG mutant was assayed for PM resistance in a PmrA^c background and the pmrG mutation was shown to have no effect. The gene downstream of the operon, pmrD, however, had previously been shown to have an effect on PM resistance only in high copy number. To see if it too was regulated by PmrA-PmrB, a firefly luciferase fusion was created and expression was measured when strains were grown in various backgrounds. This data surprisingly demonstrated that pmrD was regulated by PhoP-PhoQ and not PmrA-PmrB. Interestingly, work by others accomplished concurrently with our studies also showed that pmrD was regulated by PhoP-PhoQ and that the pmrD gene product mediates interaction of the PhoP-PhoQ system with the PmrA-PmrB system, as PmrA-regulated genes are not expressed in a pmrD mutant grown under PhoP-PhoQ-inducing conditions (low magnesium) (14). PmrD is thought to exert its effect on PmrA-PmrB through a posttranscriptional mechanism, possibly involving its effect on PmrA phosphorylation. These results likely explain why pmrD had an effect on PM resistance only when in multicopy and when strains were grown in LB medium (non-phoP-PhoQ-inducing conditions).

Loss of expression of the pmrHFIJKLM operon eliminates Ara4A addition to lipid A and PM resistance. To determine if this loss played a role in S. enterica serovar Typhimurium pathogenesis, a pmrF mutant was examined in mice and was shown to have markedly reduced virulence by the oral route but not by the intraperitoneal route. In addition, for most of the mice that did eventually die, the time from inoculation to death was much longer than for mice infected with a WT strain. A pmrA mutant showed similar results, but did not display as severe of a virulence defect as the pmrF mutant strain. This was surprising, as PmrA activates the pmrF-containing operon and therefore was expected to give similar results. These data suggest that the pmrHFIJKLM operon may be expressed in vivo by a mechanism independent of PmrA-PmrB.

At this time, it is unclear at which stage of infection the pmrHFIJKLM locus plays a role. The PmrA-PmrB system (and the PhoP-PhoQ system, which can activate transcription of pmrAB) is thought to be induced within macrophages, and if the role of the PmrA-PmrB regulatory system is to promote survival within macrophages, then a defect should have been observed in mice by both the intraperitoneal and oral routes. Therefore, the data suggest that the defect is prior to the interaction with macrophages. In vitro experiments show no defect in invasion or type III secretion with pmrA or pmrF mutants (J. S. Gunn, unpublished results). Therefore, it is intriguing to speculate that PmrA-PmrB may be activated by unknown environmental signals within the small intestine and that the PmrA-PmrB-induced LPS modifications play a role in resistance to intestinal AP or other intraintestinal antimicrobial factors.

The results of this work further characterize a highly regulated island of genes necessary for LPS modification, AP resistance, survival in mice, and two-component system interactions. Furthermore, these data demonstrate that in vivo-regulated modifications of LPS are an important part of Salmonella pathogenesis and suggest that the PhoP-PhoQ and PmrA-PmrB regulatory systems may be important at locations other than within host cells.