Cellular Microbiology of Mycoplasma canis

Abstract

Mycoplasma canis can infect many mammalian hosts but is best known as a commensal or opportunistic pathogen of dogs. The unexpected presence of M. canis in brains of dogs with idiopathic meningoencephalitis prompted new in vitro studies to help fill the void of basic knowledge about the organism's candidate virulence factors, the host responses that it elicits, and its potential roles in pathogenesis. Secretion of reactive oxygen species and sialidase varied quantitatively (P < 0.01) among strains of M. canis isolated from canine brain tissue or mucosal surfaces. All strains colonized the surface of canine MDCK epithelial and DH82 histiocyte cells and murine C8-D1A astrocytes. Transit through MDCK and DH82 cells was demonstrated by gentamicin protection assays and three-dimensional immunofluorescence imaging. Strains further varied (P < 0.01) in the extents to which they influenced the secretion of tumor necrosis factor alpha (TNF-α) and the neuroendocrine regulatory peptide endothelin-1 by DH82 cells. Inoculation with M. canis also decreased major histocompatibility complex class II (MHC-II) antigen expression by DH82 cells (P < 0.01), while secretion of gamma interferon (IFN-γ), interleukin-6 (IL-6), interleukin-10 (IL-10), and complement factor H was unaffected. The basis for differences in the responses elicited by these strains was not obvious in their genome sequences. No acute cytopathic effects on any homogeneous cell line, or consistent patterns of M. canis polyvalent antigen distribution in canine meningoencephalitis case brain tissues, were apparent. Thus, while it is not likely a primary neuropathogen, M. canis has the capacity to influence meningoencephalitis through complex interactions within the multicellular and neurochemical in vivo milieu.

INTRODUCTION

Mycoplasma canis infects many mammalian hosts but is usually thought of as a commensal or opportunistic cofactor in respiratory or urogenital tract diseases of dogs (1). We found unexpectedly that M. canis was also detectable by culture or PCR in a majority of brain tissue specimens in a retrospective case-control study of canine granulomatous meningoencephalitis (ME) (GME) and necrotizing ME (NME) (2). The presence of M. canis in brain tissue was associated with both GME and NME (both P < 0.05, as determined by a χ² test). The clinical signs of this common idiopathic neurological disease of dogs include seizures, proprioceptive deficits, circling, and blindness. Immunosuppressive therapy may be palliative, but the syndrome is progressive and uniformly fatal (3). The extensive search for a presumed viral cause of canine GME and NME has been fruitless (4).

In humans, bacterial meningitis and encephalitis are multifactorial lethal infections with often severe sequelae for survivors. New detection methods have shown that the variety of bacteria associated with human ME is much more extensive than usually appreciated (5–9). Additional animal models of bacterial ME are necessary to study this broader spectrum of pathogens (10). Since a possible association between M. canis and canine ME was discovered, our objective has been to help fill the void of basic knowledge about the organism's virulence factors, the host responses that it elicits, and its potential roles in pathogenesis. Our working hypotheses were that M. canis is capable of evoking host cell responses that favor dissemination from mucosal surfaces to secondary sites of infection, possibly in a strain-dependent fashion, and also that, regardless of how it might reach those sites, the presence of M. canis there modulates inflammation and direct injury to host cells. Understanding this potential can be expected to help evaluate the cause of canine ME and other diseases.

(Portions of these data were presented in abstract form at Congresses of the International Organization for Mycoplasmology [90, 91].)

MATERIALS AND METHODS

Mycoplasma strains and cultivation.

Strain PG14^T of M. canis (ATCC 19525) was first isolated from the throat of a normal dog (11). Strains UF31, UF33, LV, 5, 26, Cal, and Mara were first isolated from vaginal swabs of dogs without ME (12). Strains UFG1, UFG2, UFG3, and UFG4 were isolated from frozen brain tissues from cases of canine NME (2). Mycoplasma cynos strain H-831^T (ATCC 27544) was first isolated from the lung of a dog with pneumonia (13). Mycoplasma arginini strain G230^T, Mycoplasma bovigenitalium strain PG11^T, Mycoplasma edwardii strain PG24^T, Mycoplasma maculosum strain Skotti B, Mycoplasma molare strain H542^T, Mycoplasma opalescens strain MH5408^T, and Mycoplasma spumans strain PG13^T, representing other species that have been isolated from dogs (1), were obtained from The Mollicutes Collection. All strains were propagated under standard conditions (14) in ATCC 988 medium supplemented with fetal bovine serum (FBS) and glucose or arginine. Stock culture density expressed in CFU was determined by serial dilution and colony counting after 5 to 7 days of incubation.

Scanning electron microscopy.

M. canis strain PG14^T cells were prepared for scanning electron microscopy (SEM) as previously described (15), with minor modifications. Briefly, glass coverslips were placed into wells of a 24-well plate. In each well, 100 μl of an M. canis stock was inoculated into 400 μl SP-4 broth supplemented with 3% gelatin. After 3 h at 37°C, coverslips were fixed for 30 min at room temperature in 1.5% glutaraldehyde–1% paraformaldehyde–0.1 M sodium cacodylate (pH 7.2), rinsed with 0.1 M sodium cacodylate (pH 7.2) five times for 10 min, and dehydrated through a series of ethanol washes from 25% to 100% ethanol. The coverslips were then critical-point dried and gold coated. Images were viewed on a Zeiss Supra 35 (full-frame E-mount, Gold series) variable-pressure scanning electron microscope operating at 4 kV.

Anti-M. canis polyclonal antibody production, purification, and validation.

Polyclonal antisera against whole-cell lysates (16) of M. canis strains PG14^T, UF33, and UFG1 were generated individually in rabbits. Antiserum production was performed by Lampire Biological Laboratories with approval of its Institutional Animal Care and Use Committee. The protocol consisted of initial subcutaneous inoculation with lysate plus complete Freund's adjuvant, followed by boosters of lysate plus incomplete Freund's adjuvant. Seroconversion and strain cross-reactivity were assessed by standard indirect enzyme-linked immunosorbent assay (ELISA) methods. All nine homologous and heterologous antigen-antiserum combinations were examined.

Total IgG was then purified from each antiserum by circulation through Sepharose-coupled protein G affinity chromatography columns according to procedures recommended by the supplier (catalog no. 17-0618-05; GE Healthcare). For each antiserum, eluate fractions with peak A₂₈₀ values were pooled, dialyzed against 1× phosphate-buffered saline (PBS), and then desalted and concentrated to a final protein concentration of 15 mg/ml. A portion of each preparation was labeled with biotin according to procedures recommended by the supplier (catalog no. PI21329, PI89889, and PI28005; Pierce) for use as immunohistochemical probes. The working dilutions of the biotinylated probes were assessed by standard direct ELISA methods. All nine homologous and heterologous antigen-probe combinations were examined.

Host cell sources and cultivation.

Canine MDCK epithelial cells, first isolated from the kidney of a normal dog (17); canine DH82 histiocytes, first isolated from a dog with malignant histiocytosis (18); and murine C8-D1A type I astrocytes, first isolated from the cerebellums of normal C57BL/6 mice (19), were freshly obtained from the ATCC (ATCC CCL-34, ATCC CRL-10389, and ATCC CRL-2541, respectively). All cell lines were cultivated as monolayers in the recommended medium supplemented with antibiotic-antimycotic (Cellgro, catalog no. 30-004-Cl; Corning) in uncoated filter cap flasks and subcultured according to methods recommended by the ATCC.

Host cell colonization.

MDCK, DH82, and C8-D1A cells at 50% confluence on polystyrene vessel, tissue culture-treated, uncoated glass chamber slides (catalog no. 354104; BD Falcon) were inoculated with live or killed (by incubation with 10 μg/ml tetracycline overnight at 37°C) strains of M. canis in 1 ml of cell culture medium; incubated under the conditions described above; and then fixed with 4% (vol/vol) paraformaldehyde in PBS for 10 min on ice, permeabilized with 0.1% (vol/vol) Triton X-100 for 10 min at 25°C, and blocked with 4% (vol/vol) FBS overnight at 4°C. A sham-inoculated negative control of serum-free Dulbecco's modified Eagle's medium (DMEM) was routinely included. The slides were probed with primary mouse monoclonal IgG1 anti-α-tubulin (catalog no. T9026; Sigma-Aldrich) at a 1:500 dilution or mouse monoclonal IgG1 clone GA5 anti-glial fibrillary acidic protein (GFAP) (catalog no. 14-9892-80; eBioscience) at a 1:500 dilution and our unbiotinylated rabbit polyclonal IgG anti-M. canis lysate at a 1:500 dilution, coincubated for 1 h at 25°C or overnight at 4°C. The secondary antibodies were Alexa 488-conjugated goat anti-mouse IgG (catalog no. A11001; Invitrogen) at a 1:2,000 dilution and Alexa 546-conjugated goat anti-rabbit IgG (catalog no. A11010; Invitrogen) at a 1:1,000 dilution, coincubated for 1 h at 25°C or overnight at 4°C. The antibodies were diluted in an antibody diluent solution (catalog no. 00-3118; Invitrogen). Controls were probed with secondary antibody only. Coverslips were attached by using 60 μl/slide of mounting medium with 4′,6-diamidino-2-phenylindole (DAPI) (catalog no. P36931; Invitrogen). Two-dimensional digital images were recorded by using an EVOS FL fluorescence microscope. Z-series image stacks were recorded at 0.2- to 0.3-μm intervals by using a Leica TCS-SP5 confocal fluorescence microscope and subjected to image deconvolution analyses in three-dimensional digital reconstructions of individual infected cells by using the Volocity Acquisition software module (Perkin-Elmer).

Preliminary time course studies were conducted to establish the rate of cell colonization in vitro by M. canis. Adherent MDCK cells in chamber slides were inoculated with strain PG14^T at an initial multiplicity of infection (MOI) of ∼10 mycoplasma CFU to 1 MDCK cell and then washed and fixed for immunofluorescence imaging after 0, 2, 6, 16, 24, and 48 h of incubation.

Stable tagging of M. canis strains PG14^T, UF31, UF33, UFG1, and UFG4 and M. cynos strain H-831^T with the monomeric red fluorescent protein mCherry was also attempted by random transposon insertion using the mini-Tn4001-tetM plasmid vector pTF20mChloxp (see the supplemental material).

Cell invasion.

The capacity of M. canis strains to invade canine cells was also assessed by using a gentamicin protection assay (20). After a cell viability count was conducted by using trypan blue staining, MDCK and DH82 cells were washed and resuspended in serum-free DMEM; seeded into duplicate 6-well tissue culture-treated polystyrene plates (Costar, catalog no. 3516; Corning); and then inoculated with washed M. canis strain PG14^T, UFG1, UF31, UF33, or LV cells at an initial mycoplasma CFU/viable host cell MOI of 10:1. After 24 or 48 h of coincubation at 37°C in 5% CO₂, the medium was exchanged with serum-free DMEM plus 400 μg/ml gentamicin. Controls received fresh medium without antibiotic. Incubation was continued for 3 h at 37°C in order to eradicate all extracellular mycoplasmas. The host cells were finally washed and resuspended in serum-free DMEM without antibiotic, and serial dilutions were then inoculated onto SP-4 agar to allow colonies to form from any M. canis cells that had evaded the effects of gentamicin by intracellular invasion. The numbers of colonies were counted after 5 to 7 days of incubation at 37°C in 5% CO₂. Each colony was interpreted to represent one invaded host cell in order to estimate the frequency of invasion.

Sialidase.

Most strains of M. canis secrete a form of sialidase (neuraminidase) encoded by the nanI gene that might modulate cytadherence, transmission, and possibly host cell injury (12). MDCK cells display terminal sialic acid with both α-(2,3) and α-(2,6) linkages to subterminal galactose in cell surface antennary oligosaccharides (21, 22), but the surface glycosylation patterns of DH82 histiocytes and C8-D1A type I astrocytes are not well documented. The specificity of the M. canis sialidase NanI for α-(2,3)- and/or α-(2,6)-linked sialic acid was determined by using the lectins Maackia amurensis agglutinin (MAA), which binds to terminal sialic acid with an α-(2,3) linkage to galactose, and Sambucus nigra agglutinin (SNA), which binds to terminal sialic acid with an α-(2,6) linkage to galactose, to assess desialylation of the glycoprotein substrates fetuin (23) and transferrin (24), respectively. Triantennary fetuin includes both α-(2,3) and α-(2,6) sialic acid linkages, while biantennary transferrin has only the α-(2,6) linkage to galactose. Briefly, 5 μg of fetuin and transferrin was incubated separately with 2.5 × 10⁸ CFU of washed M. canis PG14^T cells in 25 μl of glucose-free RPMI medium for 48 h at 37°C. Desialylation was detected in a Western blot format by using digoxigenin (DIG)-labeled lectin probes and alkaline phosphatase-conjugated antidigoxigenin polyclonal sheep Fab fragments according to the methods provided by the lectin supplier (DIG glycan differentiation kit, catalog no. 11210238001; Roche).

Atypical M. canis strain LV lacks any detectable sialidase activity (12). For the present studies, the genetic basis of its sialidase-negative phenotype was determined by whole-genome sequencing. Complete de novo assembly of Illumina GAIIx paired-end sequencing reads from the strain LV genome was performed by using a combination of Celera, Ray, and Edena softwares (25–27). Annotation was accomplished via the NCBI Prokaryotic Genome Annotation Pipeline (28). Nucleotide and amino acid sequence similarities to the genomes of previously sequenced strains PG14^T, UF31, UF33, UFG1, and UFG4 (29) were calculated by using JSpecies version 1.2.1 (30) and AAI Calculator (http://enve-omics.ce.gatech.edu/aai).

The influence of host cell surface sialylation on colonization by M. canis was assessed by using three complementary approaches. First, MDCK cell layers in chamber slides were desialylated by preincubation for 1 h at 37°C in DMEM containing 1 U/ml of exogenous sialidase purified from Clostridium perfringens (catalog no. N-3001; Sigma-Aldrich) (31). The enzyme removes α-(2,3)-, α-(2,6)-, and α-(2,8)-linked cell surface sialic acid, with the α-(2,3)-linked residues being cleaved most efficiently (32). The cells were then washed; inoculated with strain PG14^T, UF33, or LV; and imaged in z-series stacks as described above after 24 h of incubation to minimize the potential for resialylation. Uninfected controls were preincubated with 0, 2, 5, or 10 U/ml C. perfringens sialidase. Second, the endogenous sialidase secreted by M. canis strain PG14^T was inhibited by preincubation of 1 × 10⁶ CFU for 1 h at 37°C in DMEM containing 0, 0.05, 0.1, 0.5, 1, or 5 mg/ml of 2-deoxy-2,3-didehydro-N-acetylneuraminic acid (DANA) (catalog no. 252926; Calbiochem), a competitive inhibitor of bacterial, viral, and mammalian sialidases. The suspensions of M. canis in DMEM containing DANA were then transferred onto MDCK cell layers in chamber slides and imaged in 0.2-μm z-series stacks after 48 h of incubation. Uninfected controls were incubated in DMEM containing 0 or 3.3 mg/ml DANA. Colonization rates in these two studies were quantitated objectively by measuring the total number of anti-M. canis antibody-labeled voxels per cell in reconstructed three-dimensional immunofluorescence images (n = 3 random fields of view/strain, with 30 to 200 cell nuclei/field) by using a “shrink-wrapping” lasso tool to circumscribe precisely the total cellular volume defined by the cytoskeletal label (Volocity Quantitation software module; Perkin-Elmer). As a third approach, untreated MDCK, DH82, and C8-D1A cells were inoculated with sialidase-negative M. canis strain LV in chamber slides and imaged after incubation as described above for the cell invasion studies.

Reactive oxygen species production.

The potential for M. canis strains PG14^T, UF31, UF33, UFG1, and UFG4 and M. cynos strain H-831^T to produce H₂O₂ and possibly other reactive oxygen species (ROS) was documented by using an assay based on the conversion of 10-acetyl-3,7-dihydroxyphenoxazine to fluorescent resorufin in the presence of horseradish peroxidase in a 96-well-plate format. Briefly, 5 × 10⁴ CFU of each strain were pelleted by centrifugation and either resuspended immediately in a proprietary assay buffer in duplicates according to methods provided by the supplier (catalog no. STA-344-T; Cell Biolabs) or resuspended in PBS and starved for 60 min at 37°C before harvest and resuspension in assay buffer supplemented with 100 μM glycerol (33). The final reaction mixture volume was 100 μl. After an initial 30-min incubation at 25°C, the relative fluorescence at 590 nm was measured at 10-min intervals for another 60 min. The amount of ROS present was estimated by comparing the amount of resorufin accumulated in each specimen to that generated by a series of standards containing 0 to 100 μM H₂O₂.

Cytokine expression.

DH82 cells on chamber slides were inoculated with M. canis when they had reached ~50% confluence. In trial 1, cells were inoculated in duplicates with live cells of M. canis strain PG14^T, LV, UFG1, UF31, or UF33 that had been washed and resuspended in serum-free DMEM without stabilized antibiotic-antimycotic at an initial mycoplasma CFU/DH82 cell MOI of 10:1. Negative controls were sham inoculated with serum-free DMEM, and positive controls were stimulated with 10 μg/ml standard lipopolysaccharide (LPS) extracted from Escherichia coli O111:B4 with a phenol-water mixture. Culture supernatant liquid was harvested at 48 h postinoculation and immediately frozen in liquid nitrogen until assayed. In trial 2, DH82 cells were inoculated in duplicates with live or killed cells of strain PG14^T or UFG1 that had been washed and resuspended in serum-free DMEM with stabilized antibiotic-antimycotic, at initial MOI of ~10:1 or 1,000:1 for each strain. Aliquots of culture supernatant liquid were harvested at 0, 24, 48, 72, and 96 h postinoculation and immediately frozen in liquid nitrogen until assayed. Levels of gamma interferon (IFN-γ), interleukin-6 (IL-6), interleukin-10 (IL-10), and tumor necrosis factor alpha (TNF-α) were measured by using a custom Milliplex canine cytokine/chemokine panel in a 96-well-plate format according to methods provided by the supplier (catalog no. CCYTOMAG-90K; EMD Millipore) and xMAP fluorometry systems (Luminex LX200 in trial 1 and Bio-Plex MagPix in trial 2).

MHC-II antigen expression.

DH82 cells at 50% confluence in 25-cm² flasks were inoculated with live cells of M. canis strain PG14^T, LV, UFG1, UFG2, UFG3, or UFG4 under the conditions described above for cytokine trial 1, including a sham-inoculated negative control of serum-free DMEM and a positive control treated with 10 μg/ml LPS. After 48 h, the cells were harvested by using a cell scraper and then washed twice with flow wash buffer (FWB) consisting of 1 mg/ml bovine serum albumin (BSA) plus 1 mM EDTA in PBS. Fc-binding receptors specific for IgG2a (34) were blocked in FWB plus 10% (vol/vol) FBS or rat serum (catalog no. R-9759; Sigma-Aldrich) for 15 min at 4°C, and duplicates of ∼1 × 10⁶ cells were then resuspended in 90 μl of FWB plus 10 μl (0.5 μg) of Alexa 647-conjugated rat IgG2a monoclonal anti-canine major histocompatibility complex class II (MHC-II) antibody (clone YKIX334.2) (catalog no. 51-5909; eBioscience) (35) and incubated in the dark for 1 h at 25°C. Isotype controls were incubated with the same concentration of an Alexa 647-conjugated mouse (catalog no. 51-4724; eBioscience) or rat (catalog no. 557690; BD Pharmingen) IgG2aΚ monoclonal antibody against keyhole limpet hemocyanin (KLH). After three washes in FWB, the cells were resuspended in 4% (vol/vol) paraformaldehyde for 10 min at 4°C and then washed, resuspended in 250 μl of FWB, and strained through prewetted 35-μm nylon mesh into round-bottom polystyrene tubes (catalog no. REF352235; Falcon) for flow cytometry (36). The fluorescence intensity of 5 × 10⁴ to 1 × 10⁵ events (cells) was measured for each specimen.

Endothelin-1 and complement factor H expression.

DH82 cells at 50% confluence in 25-cm² flasks were inoculated with live cells of M. canis strain PG14^T, UF31, UF33, LV, or UFG1 under the conditions described above for MHC-II expression, including a sham-inoculated negative control of serum-free DMEM and a positive control of 10 μg/ml standard LPS. The culture supernatant was harvested at 48 h postinoculation and immediately frozen in liquid nitrogen until assayed. The levels of the vasoconstrictive and neuroendocrine regulatory peptide endothelin-1 (ET-1) (37, 38) and canine complement factor H were quantitated by using colorimetric sandwich ELISAs according to methods recommended by the suppliers (R&D Systems [catalog no. DET100] and NovaTeinBio [catalog no. BG-CAN10534], respectively).

Statistical analyses.

The effects of the above-described treatments on colonization rates and expression levels of cytokines, ET-1, and complement factor H were analyzed by analysis of variance (ANOVA) using JMP Pro v11.0 (SAS Institute) and by post hoc Tukey-Kramer honestly significant difference (HSD) or Dunnett comparisons to controls when main effects were significant (P < 0.05). The relative levels of MHC-II expression by infected DH82 cells were compared to those of uninoculated controls by using Kolmogorov-Smirnov and probability binning statistical algorithms (FlowJo v9.6; BD Biosciences). A χ^(T) value of >4 was considered significant (39).

Immunohistology.

Formalin-fixed, paraffin-embedded brain tissue sections from three archived cases of NME and three archived cases of GME were processed for immunohistochemical staining by using the biotinylated polyclonal anti-M. canis strain PG14^T probe described above. Controls were fixed brain tissue sections from archived canine cases of idiopathic cardiac necrosis, chronic hepatopathy, and hemangiosarcoma. The slides were deparaffinized and rehydrated by standard methods, including an intermediary step to quench endogenous peroxidase activity (3% H₂O₂ in methanol). For enzyme-induced antigen retrieval, sections were submerged in a proprietary trypsin solution (catalog no. 00-3006; Invitrogen). For heat-induced antigen retrieval, sections were heated in a steamer while being submerged in a neutral-pH EDTA solution (catalog no. 920P; Cell Marque). Slides were then coincubated with a universal blocking reagent (catalog no. BS966; Biocare Medical) and a polyclonal anti-M. canis probe at a 1:2,000, 1:4,000, or 1:8,000 dilution for 16 h at 40°C, followed by application of a peroxidase-conjugated secondary antibody (catalog no. RHRP520; Biocare Medical). Bound probe was detected by incubating slides in 3′,3′-diaminobenzidine (DAB) (catalog no. DB801; Biocare Medical). Slides were counterstained with hematoxylin, and coverslips were mounted by using an antioxidant medium (catalog no. 8312-4; Thermo Scientific). The entire surface of each slide was inspected for accumulation of oxidized DAB precipitates.

RESULTS

Genomics and transformation.

The closed circular M. canis strain LV genome assembly (GenBank accession no. CP011368.1) was 968,791 bp in length and had 27% G+C content, with 97 to 98% average nucleotide identity and 98.1 to 99.5% predicted amino acid identity to five other strains of M. canis. The tetranucleotide frequency correlation with the other strains was 0.9955 to 0.9957. This is the first complete genome of M. canis. A 1.9-kb insertion evident in the sialidase homolog nanI was confirmed by PCR-based Sanger sequencing. It consisted of an insertion sequence (IS) encoding a 509-amino-acid (aa) isoform of a streptococcal IS1202 group transposase (blastp E value of 2e⁻¹⁶) flanked by perfect 90-nucleotide (nt) direct repeats, one of which also occurs in the wild-type nanI open reading frame (ORF) of PG14^T and other strains of M. canis. This insertion introduced multiple premature stop codons into the nanI ORF of strain LV, which was sufficient to explain its sialidase-negative phenotype. The transposase had regions of very high similarity (blastp E values of 1e⁻⁴⁸ to 4e⁻⁹¹) to two series of short predicted ORFs, MYCN0660 to MYCN0662 and MYCN0767 MYCN0769, annotated in the genome of the canine pathogen M. cynos strain C142 (GenBank accession no. HF559394.1). M. cynos strain H-831^T was also PCR positive for the transposase, whereas 10 other isolates of M. canis and M. arginini strain G230^T, M. bovigenitalium strain PG11^T, M. edwardii strain PG24^T, M. maculosum strain Skotti B, M. molare strain H542^T, M. opalescens strain MH5408^T, and M. spumans strain PG13^T were PCR negative, providing evidence that the sialidase-negative phenotype of M. canis strain LV results from insertional inactivation by an IS acquired naturally from M. cynos, possibly during coinfection of a canine host. No other relevant differences from the genomes of other strains of M. canis were recognized.

All strains of M. canis, and M. cynos H-831^T, resisted stable genetic transformation with IS256-based random transposition vectors under a multitude of polyethylene glycol (PEG), Lipofectin, Lipofectamine, and electroporation conditions (see the supplemental material). However, the presence of an IS1202 group transposase isoform in the genome of strain LV suggests that M. canis may be amenable to transformation with IS1202-based constructs in the future.

Colonization and invasion.

Under standard SEM conditions, cells of strain PG14^T were coccoid, with a diameter of ∼350 nm, which is ∼1/25 of the diameter of a typical MDCK cell nucleus. They had a lightly textured surface marked by 20-nm circular elevations occurring in a pattern of stripes along the axis of the cells (see Fig. S1 in the supplemental material). No quantitative differences in binding to M. canis whole-cell lysate antigens of different strains, or qualitative differences in the labeling pattern of cells of different M. canis strains adherent to host cells in vitro, were observed among the polyclonal antibody reagents generated for these studies, so the anti-PG14^T antibody was adopted as the standard probe. In preliminary time course studies, little surface colonization of MDCK or DH82 cells was apparent microscopically before 24 h postinoculation, and the background of noncytadherent M. canis cells increased substantially beyond 48 h postinoculation, so 48 h was adopted as the standard incubation period for most subsequent experiments. For all strains examined, M. canis was most often observed attached at the margins of host cells in monolayers or in close proximity to host cell nuclei (Fig. 1A to D). Intracytoplasmic M. canis cells (Fig. 1B) were also evident in three-dimensional image reconstructions of infected host cells of all types. Complementary evidence of host cell invasion was the recovery of viable cells of strains PG14^T, UF31, and UF33 as early as 24 h postinoculation from both MDCK and DH82 cells subsequently treated with gentamicin (Table 1). Cells of strains PG14^T, UF31, and UF33 were recovered from the interior of about 1 of every 10 host cells inoculated, while cells of strains LV and UFG1 were not recovered from either MDCK or DH82 cells following gentamicin treatment. The remains of killed M. canis cells also bound extensively to host cells (see Fig. S2 in the supplemental material).

FIG 1.

Cell colonization and invasion by M. canis PG14^T. (A and B) Crosshairs indicate perinuclear (A) and intracytoplasmic (B) immunolabeled M. canis cells (red) in MDCK epithelial cells optically sectioned in the z-axis at 0.2-μm intervals. The cytoskeleton was labeled with anti-α-tubulin (green), and nuclei were labeled with DAPI (blue). (C and D) Similar patterns of colonization were observed with DH82 histiocytes labeled with anti-α-tubulin (C) and with C8-D1A astrocytes labeled with anti-glial fibrillary acidic protein (D).

TABLE 1.

Intracellular invasion and persistence of Mycoplasma canis in cultured canine cells^a

Strain	Mean CFU in canine cells
	− gentamicin				+ gentamicin
	24 h		48 h		24 h		48 h
	MDCK	DH82	MDCK	DH82	MDCK	DH82	MDCK	DH82
PG14T	1.1 × 103	7.0 × 102	1.5 × 103	8.8 × 102	0	0	6.3 × 102	8.5 × 102
UFG1	1.5 × 104	1.5 × 104	1.6 × 104	1.5 × 104	0	0	0	0
LV	8.8 × 102	8.5 × 102	9.4 × 103	3.3 × 103	0	0	0	0
UF31	9.9 × 103	1.6 × 104	1.1 × 104	2.0 × 104	4.1 × 103	8.6 × 102	4.3 × 103	4.9 × 103
UF33	2.1 × 104	2.0 × 104	2.2 × 104	2.4 × 104	3.6 × 103	1.1 × 103	3.8 × 103	3.3 × 103
Pooled SEM	3.9 × 103	4.1 × 103	3.4 × 103	4.6 × 103	3.0 × 102	1.0 × 102	1.1 × 103	1.2 × 103

^aValues are mean CFU (n = 2) recovered from 1 × 10⁵ viable nonphagocytic MDCK epithelial or phagocytic DH82 histiocyte cells in serum-free medium inoculated with washed M. canis cells at an MOI of 10. One CFU represents one infected host cell. Times shown are hours of coincubation of M. canis with canine cells before exposure to 400 μg/ml gentamicin for 3 h at 37°C to eradicate extracellular mycoplasmas.

Role of sialidase in colonization.

Sialidase-negative strain LV colonized untreated MDCK cells as efficiently as wild-type strain PG14^T did, and quantitative strain differences (P < 0.05) in the extents of colonization either with or without prior desialylation were at most 2-fold (Fig. 2A). However, pretreatment with exogenous C. perfringens sialidase to desialylate the surfaces of MDCK cells significantly reduced (P < 0.01) their colonization by all strains tested 24 h after inoculation (Fig. 2A). Furthermore, coincubation of strain PG14^T with DANA, a competitive inhibitor of its endogenous sialidase, significantly enhanced (P < 0.05) the number of M. canis cells remaining attached to untreated MDCK cells 48 h after inoculation in a concentration-dependent fashion (Fig. 2B). The size, shape, and number of adherent uninfected control cells were unaffected by the above-described treatments with C. perfringens sialidase or DANA.

FIG 2.

Role of sialidase in M. canis colonization. (A) Pretreatment of MDCK cells with exogenous sialidase reduces the extent of colonization by all strains tested. Endogenous sialidase-negative strain LV colonized as efficiently as wild-type strain PG14^T did. (B) Coincubation with DANA, a competitive inhibitor of sialidase, enhances the number of M. canis cells remaining attached to MDCK cells 48 h after inoculation in a concentration-dependent fashion. Means (± standard errors of the means) with different superscripts differ (P < 0.05 by Tukey-Kramer HSD post hoc comparisons). (C) Binding of MAA lectin to fetuin is completely abolished by incubation of the ligand with M. canis, showing that the secreted sialidase of M. canis efficiently removes terminal sialic acid linked with α-(2,3) to galactose. Comparatively less transferrin was modified by incubation with M. canis, as evidenced by the persistence of SNA lectin binding to a predominant 65-kDa band on Western blots and by the appearance of a new 45-kDa band indicating the removal of only one of transferrin's two α-(2,6)-linked terminal sialic acid residues. The blots were spliced for labeling purposes.

The secreted sialidase of M. canis was shown to have a preference for terminal sialic acid with an α-(2,3) linkage to subterminal galactose when binding of MAA lectin to fetuin was completely abolished by incubation of the ligand with M. canis cells (Fig. 2C). Less transferrin was modified by incubation with M. canis, as evidenced by the persistence of SNA binding to a predominant 65-kDa band on Western blots and by the appearance of a new 45-kDa band indicating the removal of only one of transferrin's two terminal sialic acid residues with an α-(2,6) linkage to subterminal galactose. M. canis NanI therefore had the same linkage specificity and relative cleavage efficiency as those of the sialidase purified from C. perfringens.

Reactive oxygen species.

The rate of ROS excretion was strain variable, with the majority of nonstarved cells of the M. canis strains generating a minor increase in excretion in the assay medium, equivalent to <2 μM H₂O₂/10⁵ CFU/h, although a few strains, including brain isolate G1, generated 3- to 5-fold more ROS (P < 0.05) (Fig. 3). Endogenous glycerol-3-phosphate was the presumed substrate for glycerol-3-phosphate oxidase-dependent H₂O₂ production under these conditions, because following starvation before the assay, no strain of M. canis excreted detectable ROS into glycerol-supplemented assay medium (data not shown).

FIG 3.

Strain-variable excretion of ROS by freshly harvested M. canis cells in glycerol-free assay medium. The rate of excretion of H₂O₂ and possibly other ROS was measured by using an assay based on the oxidation of 10-acetyl-3,7-dihydroxyphenoxazine to fluorescent resorufin. Means (± standard errors of the means) with different superscripts differ (P < 0.05 by Tukey-Kramer HSD post hoc comparisons). Following starvation for 1 h before the assay, no strain excreted detectable ROS in glycerol-supplemented assay medium (data not shown). M. cyn, Mycoplasma cynos strain H-831^T.

Cellular responses to colonization.

When measured at 48 h postinoculation, strains of M. canis varied significantly (P < 0.01) in the extent to which they influenced the in vitro secretion of TNF-α by DH82 histiocytes, while the effect of strains did not reach statistical significance for the secretion of IFN-γ, IL-6, or IL-10 (Fig. 4A to D). Vaginal isolates UF31 and UF33 evoked a Th1-type proinflammatory TNF-α, IL-6, and IFN-γ profile, whereas PG14^T, LV, and especially brain isolate UFG1, in contrast, seemed anti-inflammatory, with high IL-10/IFN-γ ratios and the smallest amount of TNF-α elicited. The concentration of IFN-γ in the cell culture medium did not increase (P > 0.05) beyond 48 h of incubation with strain PG14^T or UFG1, but TNF-α, IL-6, and IL-10 concentrations continued to increase steadily (P < 0.01) through 96 h. A similar time course of responses for TNF-α, IL-6, and IL-10 was observed when the MOI was increased to 1,000:1, although the concentrations of all cytokines measured were unexpectedly lower (P < 0.01) at the higher MOI. Exposure of DH82 cells to killed cells of strain PG14^T or UFG1 had a tendency to induce slightly more (P < 0.10) IL-10 at each time point than inoculation with live M. canis cells did, and killed cells elicited the secretion of as much of the other cytokines as live M. canis cells did.

FIG 4.

Cytokine responses of DH82 histiocytes to colonization with M. canis. Data shown are from 48 h after inoculation with live M. canis cells. M, untreated controls; LPS, controls incubated with 10 μg/ml E. coli LPS. Means (± standard errors of the means) with different superscripts differ (P < 0.05 by Tukey-Kramer HSD post hoc comparisons).

Histiocyte MHC-II antigen expression was uniformly suppressed (P < 0.01) by all strains, with 40 to 60% decreases in the median fluorescence intensity induced by strains PG14^T and LV (Fig. 5) and 15 to 45% decreases induced by the other strains tested. The median fluorescence intensity of uninoculated controls incubated in 10 μg/ml LPS increased 12% (P < 0.01). No effects of the type of blocking serum (bovine versus rat) or amount of primary antibody (0.05 to 1.25 μg/10⁶ DH82 cells) were detected, but uninoculated histiocytes exposed to the mouse isotype control monoclonal antibody had consistently greater nonspecific fluorescence than did those exposed to the rat isotype equivalent.

FIG 5.

Effects of M. canis on MHC-II antigen expression by DH82 histiocytes. Decreases in the median fluorescence intensity of 40 to 60% were induced by strains PG14^T and LV, while 15 to 45% decreases were induced by the other strains tested [P < 0.01 by χ^(T) test].

Secretion of ET-1 by DH82 cells was inversely related to TNF-α and IFN-γ production, with the greatest decrease in ET-1 (P < 0.05) being elicited by strain UFG1 and only slight decreases being elicited by UF31 and UF33 (Fig. 6). The secretion of complement factor H (3.43 ± 0.05 ng/ml of medium after 48 h for 8 untreated controls) was not significantly affected by inoculation of DH82 cells with any strain of M. canis or by incubation in 10 μg/ml LPS (range, 0.5 to 4.0 ng/ml of medium).

FIG 6.

Strain-variable effect of M. canis on secretion of endothelin-1 by DH82 histiocytes. M, untreated controls; LPS, controls incubated with 10 μg/ml E. coli LPS. Means (± standard errors of the means) with different superscripts differ (P < 0.05 by Tukey-Kramer HSD post hoc comparisons).

No in vitro cytopathic effects of colonization (rounding, blebbing, vacuolization, or pycnosis) were observed under any conditions for MDCK, DH82, or C8-D1A cells inoculated with any strain of M. canis (Fig. 1).

Immunohistology.

Preliminary studies of NME lesion positive-control serial sections showed that neither trypsin nor heated EDTA antigen retrieval treatment affected the DAB staining patterns. The final method adopted employed no antigen retrieval and a probe dilution of 1:8,000 (1 μg/ml of biotinylated polyclonal anti-M. canis strain PG14^T probe). Oxidized DAB precipitates were never seen in negative-control brain specimens (Fig. 7A), but a focal diffuse precipitate was occasionally evident in case specimens (Fig. 7B and C).

FIG 7.

Immunohistochemical staining of archived NME and GME brain tissue samples probed with polyclonal anti-M. canis strain PG14^T antibody. Bound probe was detected with a peroxidase-conjugated secondary antibody and the DAB substrate. Slides were counterstained with hematoxylin (violet). (A) Negative-control canine brain parenchyma (magnification, ×10); (B) focal deposition of oxidized DAB precipitate (brown) throughout a vascular cuff in NME tissue (magnification, ×10); (C) detail from panel B (magnification, ×40).

DISCUSSION

The adult human mortality rate of 20 to 30% due to bacterial meningoencephalitis has not been significantly reduced by modern intensive care, diagnostics, or therapy (40). Sequelae among survivors include paresis, epileptic seizures, cerebral palsy, deafness, and cognitive deficits. Gaps in understanding bacterial ME reflect the prevailing research and clinical emphasis on the classic meningitis pathogens Haemophilus, Streptococcus, and Neisseria (41, 42). Fatality rates, however, are highest among the cases ultimately attributed to atypical agents, including species of Mycoplasma and other slow-growing fastidious organisms that resist standard empirical treatment with cell wall-targeting antibiotics (10, 43–46). Bacterial neuropathogens vary in their mechanisms of virulence, but assumptions regarding the importance of phagocytosis/exocytosis, fimbriae, capsules, peptidoglycan, and LPS for epithelial penetration and development of bacteremia associated with ME (47–50) are all invalid regarding the pathogenesis of mycoplasmal diseases (51, 52). This is important because the diagnostic uncertainty and empirical treatment failures that precede the adverse outcomes of many atypical infections cannot be expected to improve without more comprehensive knowledge of the etiologic agents involved (6, 44, 53, 54).

Several strains of M. canis readily colonized the surfaces of cells integral to epithelial and innate immune barriers to dissemination, as well as one type of brain cell (Fig. 1). M. canis adheres to sialylated receptors on host cells (55). The extent of initial colonization was reduced by pretreatment to remove α-(2,3)- and α-(2,6)-linked sialic acid from host cell surfaces, but the sialidase secreted by M. canis itself did not limit cytadherence to untreated cells, as shown by the similarity in adherences of wild-type strain PG14^T and sialidase-negative strain LV (Fig. 2A). The enhanced amount of cells of PG14^T remaining attached to untreated MDCK cells after incubation in the presence of the sialidase inhibitor DANA (Fig. 2B) is evidence that its sialidase may instead be important primarily for detachment from colonized host cells and subsequent transmission. The enzyme's preference for α-(2,3)-linked substrates (Fig. 2C) supports the conclusion that the enzyme adheres predominantly to α-(2,6)-linked sialoreceptors. The proportional expression of this glycosylation pattern at different anatomical sites, as well as strain differences in the amounts of sialidase secreted (12), could influence host species range, primary colonization of respiratory or urogenital tract mucosal surfaces, and dissemination and secondary localization of M. canis cells in an infected host.

The presence of visible M. canis inside nonphagocytic cells and its recovery following eradication of extracellular bacteria from both nonphagocytic and phagocytic cells demonstrated its innate capacity for intracellular invasion and persistence (Fig. 1A and B and Table 1). This finding established the potential for direct penetration through adjacent tissues, versus strictly “Trojan horse” spread via infected macrophages (56, 57), as a basis for the dissemination of M. canis from mucosal surfaces. However, the absence of cytopathic effects on any type of host cell colonized in vitro was evidence against direct injury by M. canis and was in favor of multifactorial host responses to colonization as the potential basis for any role in ME or other diseases.

In canine NME, pathological findings consist of neuroparenchymal infiltration of lymphocytes and monocytes or histiocytes, with malacia and necrosis usually being localized in the cerebral cortex and subcortical region (58–61). Similar pathologies have been attributed to other species of mycoplasmas that gained entry into the brain (62–64). The profile of cytokine secretion by DH82 cells following exposure to M. canis isolates UF31 and UF33, characterized by high TNF-α/IL-10 ratios, clearly contrasted with the responses to strains PG14^T, LV, and UFG1, characterized by high IL-10/IFN-γ ratios (Fig. 4A to D). This showed that M. canis is capable of eliciting either pro- or anti-inflammatory host responses in a strain-dependent fashion, which might influence the network of immune barriers to dissemination and recruitment and modulation of inflammatory cells at secondary sites (65). Interestingly, exposure to killed M. canis cells elicited cytokine secretion as effectively as exposure to live M. canis cells did, suggesting that diacylated membrane lipopeptides are the activating factors (66–71). The principal pathogenic effect of exposure to individual mycoplasmal lipoproteins characterized to date is transient inflammation (72), marked by local infiltration of granulocytes, macrophages, and lymphocytes; the production of proinflammatory cytokines; and complement activation (73–75).

The consistently reduced expression of MHC-II observed (Fig. 5) is evidence that the presentation of M. canis or other antigens to CD4⁺ lymphocytes could be locally compromised during mycoplasmal dissemination or by M. canis at secondary sites of infection. This effect on cellular immunity would further contribute to general immune dysregulation by strains, such as brain isolate UFG1, that elicit comparatively low IFN-γ or TNF-α responses. This pattern is consistent with DH82 cell MHC-II receptor downregulation caused by infection with the obligate intracellular canine pathogen Ehrlichia canis (36) and with the paucity of MHC-II-expressing macrophages in necrotic joint lesions of calves infected with Mycoplasma bovis by intra-articular challenge (76) but contrasts with the marked upregulation of MHC-II-expressing airway epithelial cells following intranasal inoculation of rats with Mycoplasma pulmonis (77).

ET-1 is a potent vasoconstrictor and also regulates central autonomic control of circulation, respiration, sympathetic vasomotor discharges, and the hypothalamic-pituitary axis (78). M. canis significantly reduced DH82 cell expression of ET-1 (Fig. 6), in contrast to the increases typically resulting from exposure to other bacteria (37, 79–81). Once more, the effect was greatest for brain isolate UFG1. This finding showed yet again how M. canis might influence ME in ways that differ from those of classic neuropathogens (82). For example, increased ET-1 levels during acute pneumococcal sepsis promote meningitis through vasoconstriction, resulting in cerebral ischemia (81, 83). Brain endothelial cells, microglia, astrocytes, and neurons are other sources of ET-1 (78, 84). We observed endogenous secretion levels comparable to those found in DH82 cells by Divino et al. (37), but we did not observe the reported stimulation with LPS. Methodological differences from that study included DH82 passage levels (85), the preparation of LPS (E. coli O111:B4 versus O127:B8), and endpoint after stimulation (48 h versus 24 h or less).

Canine GME is distinguished from NME by perivascular cuffing with lymphocytes, macrophages, and neutrophils as well as granulomatous lesions mainly in the cerebellum and brainstem, but both NME and GME are believed to share a pathogenesis (61). Although it did not occur frequently or consistently among cases of GME or NME, deposition of oxidized DAB throughout some vascular cuffs probed with anti-M. canis antibody (Fig. 7B and C) was evidence that the presence of M. canis in canine brains cannot be attributed solely to contamination during nonaseptic tissue collection (2). Because brain tissue specimens are typically not available for examination until long after clinical signs of chronic progressive neurological disease have developed, the abundance of intact M. canis cells remaining to be visualized at necropsy may be very low. A different polyclonal anti-M. canis antibody did not detect intact M. canis cells in similar cases examined (2).

In conclusion, no acute cytopathic effects of in vitro colonization of any host cell type or consistent patterns of M. canis polyvalent antigen distribution in canine ME case brain tissues were evident in the present studies. A causal role of M. canis in canine ME thus remains to be demonstrated. This work was limited by the paucity of commercially available canine-specific cell lines and analytical reagents and by the range of MOIs, analytes, and endpoints that were practical to sample in the in vitro infection studies. The specimens for immunohistology were not matched with unfixed tissues that could be tested by culture or PCR, and they represented only a tiny fraction of each brain. Any influence of M. canis on ME probably involves interactions within the complex multicellular and neurochemical milieu in vivo (86), which could be explored in an organotypic tissue explant model (87–89), but a prospective survey for M. canis in aseptically sampled cases of ME is needed. Positive impacts can be expected to include not only a better understanding of the ordinarily commensal organism M. canis and its role in canine diseases but also further insights into processes of bacterial invasion of the central nervous system (CNS) and subsequent injury, which is necessary to develop more comprehensive diagnostics and therapeutic interventions to reduce the burden of human and veterinary infectious neurological disease.