Diversity of Expressed vlhA Adhesin Sequences and Intermediate Hemagglutination Phenotypes in Mycoplasma synoviae

Abstract

A reservoir of pseudogene alleles encoding the primary adhesin VlhA occurs in the avian pathogen Mycoplasma synoviae. Recombination between this reservoir and its single expression site was predicted to result in lineages of M. synoviae that each express a different vlhA allele as a consequence of host immune responses to those antigens. Such interstrain diversity at the vlhA expression site, including major differences in the predicted secondary structures of their expressed adhesins, was confirmed in 14 specimens of M. synoviae. Corresponding functional differences in the extent to which they agglutinated erythrocytes, a quantitative proxy for VlhA-mediated cytadherence, were also evident. There was a >20-fold difference between the highest- and lowest-agglutinating strains and a rheostatic distribution of intermediate phenotypes among the others (Tukey-Kramer honestly significant difference [HSD], P < 0.001). Coincubation with the sialic acid analog 2-deoxy-2,3-didehydro-N-acetylneuraminate inhibited hemagglutination in a pattern correlated with endogenous sialidase activity (r = 0.91, P < 0.001), although not consistently to the same extent that erythrocyte pretreatment with sialidase purified from Clostridium perfringens did (P < 0.05). The striking correlation between the ranked hemagglutination and endogenous sialidase activities of these strains (Spearman's r = 0.874, P < 0.001) is evidence that host-induced vlhA allele switching indirectly drives sequence diversity in the passenger sialidase gene of M. synoviae.

INTRODUCTION

Mycoplasma synoviae is a pathogen associated with osteoarthritis, synovitis, and respiratory tract lesions of poultry (15, 19, 20). Cytadherence mediated by its primary adhesin VlhA is a precursor to virulence (30). Posttranslational cleavage of full-length VlhA produces the peptides MSPA (carboxy-terminal portion of VlhA) and MSPB (amino-terminal portion) (Fig. 1A). Receptor binding and cytadherence are attributed to MSPA, while the function of MSPB remains undefined (30). Interrupted expression of the vlhA gene or pretreatment with antisera against MSPA resulted in the inability of M. synoviae colonies to adsorb erythrocytes, a proxy for VlhA-mediated in vivo cytadherence. Strains able to agglutinate erythrocytes in vitro caused synovitis at a significantly higher frequency than did variants incapable of hemagglutination (29).

Fig. 1.

Sequence diversity of expressed VlhA adhesins in Mycoplasma synoviae. (A) Schematic of the vlhA translation product and its posttranslational cleavage products MSPA and MSPB. (B) Single-nucleotide substitution frequencies in the conserved and semivariable regions of M. synoviae vlhA with respect to the consensus of specimens aligned using ClustalW2. (C) Dissimilarity by blocks of 20 amino acids in the VlhA hypervariable region of the aligned specimens. This region, while having a low overall degree of similarity, includes short stretches of 70 to 100% identity. The sites and lengths of insertions (+) or deletions (−) are indicated for strains that were unique with respect to the consensus sequence.

Transcription of vlhA occurs at a single promoter next to a large assemblage of promoterless vlhA pseudogenes constituting a 69-kb locus in the M. synoviae genome. VlhA variants result from unidirectional site-specific recombination of pseudogenes into one of five specific residues at the expression site. Sequential integration of different pseudogene sequences can further result in the formation of chimeric alleles. The selective pressure of anti-VlhA antibodies in vivo is proposed to perpetuate lineages of M. synoviae that each express a different allele of vlhA as an antigen-diversifying mechanism to evade the adaptive immune system of the host (1, 30, 31, 32).

Although the specific receptor(s) engaged by VlhA remains unknown, cytadherence is mediated by sialylated moieties on the host cell surface. It has been known for a long time that receptor desialylation reduces or abolishes cytadherence by M. synoviae (22), but it was only recently discovered that adjacent to the vlhA locus, M. synoviae encodes an extracellular sialidase whose specific activity differs quantitatively among strains in a stable pattern correlated with strain virulence (26). The functional relationship between VlhA and sialidase in M. synoviae thus seems analogous to the receptor-binding hemagglutinin and receptor-destroying sialidase (neuraminidase) activities of influenza viruses. To test the prediction that sialidase activity correlates with adhesin affinity in this species (24, 44), we examined the diversity in primary amino acid sequences and predicted secondary structures of the expressed VlhAs among 14 specimens of M. synoviae and the consequent hemagglutination phenotypes of specimens with differing sialidase activities.

MATERIALS AND METHODS

Mycoplasma synoviae strains and culture conditions.

Mycoplasma synoviae strains F10-2AS, FMT, K3344, K4907A, K5016, K5395, MS117, MS173, MS178, and WVU1853^T were cultured as previously described (26). All experiments were conducted using freshly filter-cloned specimens. A medium-passage-number subclone of strain FMT (FMTp37) and a high-passage-number subclone descended from it through 90 nonselective broth passages (FMTp127 [41]) were used. Two lineages of strain WVU1853^T were examined, one being an unpassaged specimen obtained freshly from ATCC for this study (catalog number 25204, lot number 5036530; named WVUCC) and the second being a stock of ATCC 25204 passaged more than 30 times (named WVUFL [26]). The field strain 53 (MS53 [42]) and previously characterized lineages of strain WVU1853^T from Australia (“low-passage” lineage of Morrow et al. [28]; named WVUAU) and Tunisia (ATCC 25204, ca. passage 12 [18]; named WVUTU) were not readily available for phenotypic analyses, but the sequences at their vlhA expression sites (GenBank accession numbers NC_007294, AY639900, and FJ890931, respectively) were included in genotypic analyses.

Nucleotide and protein sequence analyses.

PCR primers complementary to the conserved 5′ end of the vlhA expression site (nucleotides [nt] 25 to 44; 5′-CTA TTA GCA GCT AGT GCA GT-3′) and to the flanking gapA gene (5′-CGT AAT ACT AGA CGA CCA AT-3′) were designed from the strain MS53 genome (GenBank accession no. AE017245), and the site was amplified from each specimen by standard methods. Amplicons were sequenced directly by primer walking. Sequence alignments and determination of consensus sequences were performed using Sequencher 4.7 (Gene Codes) and ClustalW2 (40). Bayesian maximum-likelihood methods were used to analyze codon variation in nine expressed vlhA sequences alignable by ClustalW2 (Selecton v2.4 software suite). The MSPA and MSPB region sequences were trimmed for this purpose in accordance with the predicted posttranslational cleavage at residue 344 of WVUAU (GenBank accession no. AAT58038). Secondary structure predictions and calculations of the percentages of residues that are part of helices, coiled coils, or β-sheets were made separately for MSPA and MSPB using GOR IV (11). Prediction of nonclassical protein secretion was performed using SecretomeP 2.0 (5).

Quantitative hemagglutination assay.

Stationary-phase cultures of M. synoviae strains F10-2AS, FMTp37, K4907A, K5016, K5395, MS117, MS173, and MS178 and lineages WVUCC and WVUFL were normalized to a final optical density at 600 nm (OD₆₀₀) of 0.05 (≈10⁷ CFU/ml), and then a 40-ml aliquot of each culture was pelleted by centrifugation. Chicken erythrocytes (Lampire Biological Laboratories) were prepared as a 15% (vol/vol) suspension in phosphate-buffered saline (PBS). Each mycoplasma pellet was resuspended in 50 μl of the erythrocyte suspension and then incubated in a round-bottom microtiter plate for 3 h at 37°C to allow hemagglutination to plateau. Agglutinated erythrocytes remained dispersed in suspension while unagglutinated ones settled to the bottom of the well. Following incubation, 10-μl aliquots were removed from the agglutinated phase, diluted 1:50 in PBS, and disaggregated by forceful passage through a tuberculin syringe fitted with a 1.5-cm, 26-gauge needle. The extent of hemagglutination was proportional to the OD₅₈₅ of the disaggregated cells. The positive agglutination control was WVUCC (21), and the negative control was PBS. To determine the extent to which hemagglutination depended on sialylation, erythrocyte suspensions either were pretreated by incubation for 1 h at 37°C with 10 U/ml of sialidase purified from Clostridium perfringens (Sigma-Aldrich) before being washed and incubated with M. synoviae or were incubated with M. synoviae in the presence of 250 mg/ml of the sialic acid analog and sialidase competitive inhibitor 2-deoxy-2,3-didehydro-N-acetylneuraminic acid (DANA; Calbiochem/EMD Biosciences).

Statistical procedures.

Both the M8 and the mechanistic-empirical combined (MEC) Bayesian statistical models of sequence evolution (38) were used to interpret the rates of nonsynonymous (amino acid-altering; K_a) and synonymous (K_s) codon usage in the aligned vlhA sequences. The M8 model uses maximum-likelihood methods accounting for factors such as different probabilities for transitions and transversions, codon bias, and among-site rate variation affecting the K_a/K_s ratio (ω) to calculate the proportion of codons with ω < 1 and the proportion with ω ≥ 1. The forms of selection acting on the sequences were interpreted from the differences in ω for each codon as calculated by the M8 model versus a null model (M8a) having a fixed maximum ω value of 1 at each residue. These differences are expected to follow a χ² frequency distribution if only neutral or stabilizing selection occurs; significant deviations from a χ² distribution indicate diversifying selection (39). The MEC model further accounts for differing empirical amino acid mutation probabilities: a position with physicochemically radical amino acid diversity is assigned a higher site-specific K_a value than is a position exhibiting only subtle variations. Compared to the more conservative M8 model, a smaller proportion of sites with ω > 1 may be sufficient for the MEC model to detect diversifying selection (see reference 25 and references therein). In addition, 90% confidence intervals were determined from the observed distribution of ω for each codon. If the lower bound of a confidence interval was >1, then the inference of diversifying selection at that residue was considered to be biologically meaningful (38, 39).

The effect of M. synoviae strain on hemagglutination (n = 3 independent replications each) was analyzed by analysis of variance (ANOVA), and Tukey-Kramer honestly significant difference (HSD) post hoc comparisons were used to group the means when the main effect was significant. The effects of pretreatment with sialidase and coincubation with DANA on hemagglutination were analyzed by ANOVA, and by Tukey-Kramer HSD or Dunnett's comparisons to the PBS control post hoc when the main effect was significant. An association between hemagglutination group rank and the ranked endogenous sialidase activities of these strains previously reported by May et al. (26) was tested by nonparametric Spearman's rank order correlation analysis. The endogenous sialidase activity of WVUCC was determined to be (1.1 ± 0.2) × 10⁻⁶ units/CFU in the spectrofluorometric assay (26). The statistical analyses were performed using JMP v7.02 (SAS). P values of <0.05 were considered statistically significant.

RESULTS

Diversity of expressed vlhA sequences.

It was possible to amplify a PCR fragment spanning the vlhA expression site from M. synoviae strains F10-2AS, FMT, K4907A, K5016, K5395, MS117, MS173, and MS178 and WVUCC and WVUFL lineages of the type strain. The amplicons ranged in size from 450 to 2,500 bp (GenBank accession numbers HQ326476 to HQ326484 and HQ682228). The vlhA expression site could not be amplified from virulent field strain K3344 (10) utilizing either those primers or two alternative primer pairs.

The vlhA allele expressed by subclone FMTp37 was identical to that of subclone FMTp127. In contrast, four lineages of WVU1853^T each expressed a different allele. The allele expressed by WVUCC encoded only an MSPA region, which was identifiable from BLASTP (but not ClustalW2) alignments supported by the presence of certain highly conserved amino acid residues (ω ≪ 1). The full vlhA sequences expressed by strain FMT and WVUTU were so divergent that they could not be objectively aligned with those of any other lineage, although the putative posttranslational cleavage site (4) was evident.

Numerous insertions, deletions, and point mutations were observed with respect to the consensus of 10 alignable vlhA sequences (including strain MS53 [GenBank accession no. YP_278398]). Nucleotide and consequent amino acid diversity occurred throughout the previously described conserved, semivariable, and hypervariable regions (2, 32). The mean (±standard error of the mean [SEM]) number of substitutions per 100 bp was 4.06 ± 1.81 for the conserved region and 7.18 ± 3.35 for the semivariable region (Fig. 1B). The mean number of substitutions per 100 bp across both regions ranged from 3.16 ± 2.25 for WVUFL to 7.3 ± 4.18 for MS53. A very high degree of diversity existed within the hypervariable region, although certain conserved sequences were also evident there (Fig. 1C).

Analyses of codon usage in the aligned sequences indicated that nonsynonymous (amino acid-altering) differences occurred more frequently than would be expected if the vlhA gene experienced either globally stabilizing or neutral forms of natural selection (P < 0.001 [38]). Only 10 codons with a lower bound of ω 90% confidence interval of >1 were detected in 344 aligned codons in MSPB, those being in the semivariable region of MSPB, but 128 of 407 aligned codons in MSPA were significantly diverse, especially throughout the first 200 amino acids downstream of the VlhA posttranslational cleavage site and most of the 50 carboxy-terminal residues in MSPA. At the posttranslational MSPB-MSPA cleavage site, the 12 residues preceding and including consensus amino acid 344 displayed 75% identity and 100% similarity across strains, while the 12 subsequent residues displayed only 8% identity and 50% similarity.

A mean of 59% of the residues in the MSPB region of the expressed VlhAs were predicted to occur in helices, while 32% and 8% were in coiled coils and β-sheets, respectively (Table 1). The MSPA region had corresponding means of 31%, 46%, and 23%, respectively. Despite substantial truncations of both MSPA and MSPB regions in the allele expressed by strain FMT, its structural composition was comparable to those of other strains. The predicted structures of MSPB expressed by strain K5395 and of MSPA expressed by strain MS53 were markedly different from those in other strains: MSPB of K5395 was dominated by residues in β-sheets rather than in helices, and MSPA of MS53 had the smallest proportion in β-sheets. The MSPA expressed by strain K5016 notably had the largest proportion in β-sheets and the smallest in coils.

Table 1.

Predicted secondary structures of VlhA alleles expressed by specimens of Mycoplasma synoviae^a

Strain	% secondary structures in region:
	MSPA			MSPB
	β-Sheets	Coiled coils	α-Helices	β-Sheets	Coiled coils	α-Helices
F10-2AS	27	47	26†	4	32	64
FMT	20	50	30	9	26†	65
K4907	25	48	27†	4	31	64
K5016	47†	25†	28	3	30	67†
K5395	27	49	24†	44†	36†	20†
MS53	16†	41†	43†	4	35†	61
MS117	19	49	32	3	30	67†
MS173	21	50	29	3	30	67†
MS178	21	50	29	3	30	67†
WVUAU	17†	48	34	7	37†	57
WVUCC	22	53†	25†
WVUFL	17†	52†	31	32†	7†	61
WVUTU	21	38	41†	5	32	60
Mean ± SEM	23 ± 2.2	46 ± 2.1	31 ± 1.6	8 ± 3.3	32 ± 1.1	59 ± 3.7
Median	21	49	29	4	31	64

^aPercentages of β-sheets, coiled coils, and α-helices in the MSPA (carboxy-terminal) and MSPB (amino-terminal) regions of full-length VlhAs, rounded to the nearest whole percents. WVUAU, low-passage-number Australian lineage; WVUCC, unpassaged lineage: WVUFL, >30-passage lineage; and WVUTU, ca. 12-passage lineage, all of the type strain WVU1853^T. The allele expressed by WVUCC encoded only an MSPA region. Superscript daggers indicate values that are >2 SEMs from the column mean.

Hemagglutination phenotypes and their correlation with endogenous sialidase activity.

A >20-fold range in quantitative hemagglutination phenotypes was observed among 12 specimens of M. synoviae tested (Tukey-Kramer HSD, P < 0.001). The WVUCC specimen agglutinated the most erythrocytes, strain K5395 did not induce hemagglutination to a significantly greater extent than PBS alone did, and there was essentially a continuous distribution of hemagglutination phenotypes among the other specimens (Table 2). No relationships between hemagglutination phenotypes and features of the secondary structures of their expressed VlhAs could be subjectively discerned (Fig. 2), but nonparametric Spearman rank correlation analysis showed that the hemagglutination phenotypes correlated positively (r = 0.874, P < 0.001) (Fig. 3) with the specific endogenous sialidase activities, which ranged >200 fold among these lineages due to single-nucleotide polymorphisms in the coding region of the sialidase gene (25, 26).

Table 2.

Hemagglutination phenotypes of Mycoplasma synoviae specimens expressing different VlhA alleles and sialidase activities^a

Strain	Hemagglutination (OD585)	HI (%) by:		Sialidase (U/CFU)c
		DANA	Sialidaseb
WVUCC	0.206 ± 0.007A	72.9 ± 3.4A	45.7 ± 19.7A	1.1 × 10−6 ± 2.0 × 10−7A
MS178	0.156 ± 0.002B	11.6 ± 5.5B	33.9 ± 1.2A	3.9 × 10−8 ± 7.8 × 10−9C
FMTp37	0.121 ± 0.001C	5.5 ± 0.9B	51.1 ± 0.6A	2.7 × 10−8 ± 1.1 × 10−8CD
FMTp127	0.119 ± 0.002C	ND	ND	3.6 × 10−8 ± 9.3 × 10−9CD
MS117	0.119 ± 0.001C	−1.1 ± 1.9B	38.9 ± 0.2A	2.1 × 10−8 ± 2.3 × 10−9DE
WVUFL	0.084 ± 0.013D	18.3 ± 14.7B	11.1 ± 16.5A	1.3 × 10−7 ± 2.7 × 10−9B
K3344	0.078 ± 0.001DE	ND	ND	1.3 × 10−8 ± 3.8 × 10−9EF
MS173	0.073 ± 0.001DE	ND	ND	2.9 × 10−9 ± 1.3 × 10−9F
F10-2AS	0.066 ± 0.004DEF	ND	ND	9.7 × 10−8 ± 1.2 × 10−9EF
K4907	0.056 ± 0.003EF	ND	ND	6.0 × 10−9 ± 3.1 × 10−9EF
K5016	0.046 ± 0.002F	−6.9 ± 4.8B	−42.5 ± 5.5B	7.2 × 10−9 ± 7.2 × 10−9EF
K5395	0.008 ± 0.002G	92.8 ± 7.1A	100C	2.7 × 10−9 ± 1.5 × 10−9F

^aValues are shown as mean ± SEMs (n = 3 independent replicates). Values within each column having different superscript capital letters are different (P < 0.05) by Tukey-Kramer post-ANOVA comparisons. Abbreviations: HI, hemagglutination inhibition (negative values indicate enhanced agglutination); DANA, coincubation with 2-deoxy-2,3-didehydro-N-acetylneuraminic acid; ND, not determined. WVUCC is an unpassaged lineage and WVUFL is a >30-passage lineage of strain WVU1853^T.

^bErythrocytes pretreated with exogenous sialidase purified from Clostridium perfringens.

^cEnzyme activity units per CFU (from the work of May et al. [26] and this report).

Fig. 2.

Predicted secondary structures of expressed full-length Mycoplasma synoviae VlhA adhesins. Long vertical bars represent α-helices, short vertical bars represent β-sheets, and horizontal bars represent coiled coils. Shading indicates the MSPB region. WVUAU, low-passage-number Australian lineage; WVUCC, unpassaged lineage; WVUFL, >30-passage lineage; and WVUTU, ca.-12-passage lineage, all of the type strain WVU1853^T. Sequences were aligned with respect to the predicted MSPB-MSPA posttranslational cleavage site of the VlhA expressed by WVUFL and are shown, beginning with WVUCC, in descending order of hemagglutination activity.

Fig. 3.

Rank correlation between hemagglutination and endogenous sialidase activity in Mycoplasma synoviae. Shading indicates the 90% confidence interval.

Modulation of hemagglutination with DANA and exogenous sialidase.

Coincubation with DANA during the hemagglutination reaction significantly inhibited hemagglutination for five of seven specimens representing the range of hemagglutinating and endogenous sialidase activities. The extent of inhibition correlated strongly with the hemagglutinating (r = 0.78, P < 0.001) and sialidase (r = 0.91, P < 0.001) activities of the specimens, but DANA did not inhibit hemagglutination to the same extent across strains as did erythrocyte pretreatment with exogenous sialidase purified from Clostridium perfringens (Table 2). For example, hemagglutination by strains FMTp37 and MS178 was inhibited by DANA (P < 0.05), but the effect was not as great as that of erythrocyte pretreatment with sialidase (P < 0.05) for those specimens. DANA was even more effective in reducing hemagglutination (P < 0.05) by WVUCC than was pretreatment with sialidase (P = 0.06), but those effects did not reach statistical significance for WVUFL. Strains FMTp37, MS117, and MS178 all showed a substantial reduction in hemagglutination following erythrocyte pretreatment with sialidase (P < 0.001), and this treatment completely abolished hemagglutination by strain K5395. Strain K5016 agglutinated more sialidase-treated than untreated erythrocytes (P < 0.05), but like strain MS117, strain K5016 hemagglutination was not significantly affected by DANA.

DISCUSSION

Although the regulatory mechanisms differ among species (46), both binary switching between “on” and “off” phases of adhesin gene transcription and more intricate switching among multiple allelic variants of an adhesin are thought to promote antigenic diversity as a basis of host immune evasion by certain pathogenic mycoplasmas (6, 12, 23, 30, 31). Adhesin allele switching is also a potential basis for systemic dissemination and variable tissue tropisms in the infected host (1, 9, 30). The switching phenomenon has been characterized at the molecular genetic level for the antigenic VlhA hemagglutinin of M. synoviae strain WVU1853^T, in which allelic variants arise through high-frequency sequential recombination between a large assemblage of promoterless adhesin pseudogenes and a single vlhA transcription site (32).

Previous analyses of intrastrain heterogeneity in the vlhA gene sequences of WVU1853^T and comparisons with other strains of M. synoviae led to the perception of conserved, semivariable, and hypervariable regions in the full-length expressed VlhAs. The conserved N-terminal region (consensus amino acids 1 to 240 of VlhA and MSPB [32]) includes a secretion signal peptide, signal peptidase recognition site, and a proline-rich repeat domain. This region, once thought to be invariably present and nearly identical among allelic variants generated by recombination (32, 33), had a rate of substitutions in the sequences that we examined that was 10-fold higher than strain variation in highly conserved genes encoding the N-acetylneuraminate degradation pathway and 100-fold higher than the 16S rRNA gene heterogeneity in M. synoviae (24). Eleven of the 30 strains sequenced by Benčina et al. (2) also were polymorphic in this region of MSPB, but it was particularly remarkable that our WVUCC specimen lacked MSPB entirely. The means by which it secretes MSPA is not evident from its expression site sequence (SecP scores of 0.83 to 0.99, when a value of >0.5 indicates possible nonclassical secretion [5]). It was previously considered that following VlhA signal peptide excision, MSPB might be anchored to the mycoplasmal cell surface via an acylated N-terminal diacylglyceryl cysteine residue (1, 4, 31), but a candidate cysteine codon occurred in only seven of the sequences that we analyzed and was also absent in a specimen derived from strain ULB925 analyzed by Benčina et al. (4). The pattern of variation in the residues flanking consensus amino acid 344 is further evidence for the precise site and high frequency of posttranslational processing, but at least one strain (K1723) fails to process VlhA due to a deletion in this region (2, 3, 33). Short segments that approached 100% nucleotide identity in the hypervariable region across strains (Fig. 1C) might serve as additional pseudogene recombination sites. The high sequence diversity downstream of consensus amino acid ∼730, and especially throughout the 50 carboxy-terminal residues of MSPA, was unexpected in the context of earlier perceptions that the sequence of this region is highly conserved (32) and speculation that hydrophobicity of this terminus is the mechanism by which MSPA associates with the mycoplasmal cell surface (4, 31).

The functional consequences of allelic sequence heterogeneity with regard to cytadherence have been interpreted only categorically as adherent versus nonadherent phase variants of M. synoviae (4, 14, 33). Noormohammadi (30) calculated that M. synoviae is capable of generating at least 10⁵ allelic variants of vlhA through recombination. The proportions of adherent and nonadherent phenotypes vary within and among strains over time at rates that seem inconsistent when interpreted categorically, probably because many vlhA alleles can confer an adherent phenotype, and there are many ways that nonadherent variants can revert to an adherent genotype (3, 4, 12, 23). For example, from an initially polymorphic culture of strain WVU1853^T, Rhoades (35) established lineages that stably maintained their subjective adherence phenotypes during 20 subsequent nonselective in vitro passages, whereas Noormohammadi et al. (32, 33) found that phase variation of strain WVU1853^T occurred so frequently in vitro that it limited their attempts to define recombination events precisely. We found quantitatively different agglutination phenotypes for filter-cloned lineages of WVU1835^T expressing different vlhA alleles and a continuous distribution of phenotypes among other filter-cloned strains, each expressing a distinct variant of vlhA. This is objective evidence of a far greater capacity for rheostasis in VlhA-mediated cytadherence than has been previously documented and of the potential for finely tuned coordination between cytadherence and other factors whose activity also differs quantitatively among strains.

From the strong correlation between agglutination and inhibition by DANA, we conclude that differences in adherence reflect avidity of the expressed VlhAs for sialic acid. This was also evidence that interstrain differences in hemagglutination were due principally to VlhA variation and not to unrecognized variation at other genetic loci. Hemagglutination inhibition by DANA was not attributable to inhibition of the endogenous sialidase activity (26), because erythrocyte desialylation with exogenous sialidase also inhibited agglutination for all specimens except K5016, which had a singular distribution of residues in β-sheets and coiled coils of MSPA. That strain may have engaged an alternative ligand (44), although a prozone-like effect on agglutination of cells treated with various concentrations of sialidase has been reported for certain other mycoplasmas (22). Regulation of binding affinity via desialylation of MSPA itself (34) cannot be ruled out because the glycosylation state of VlhA remains uncharacterized, but it seems unlikely since the 60 carboxy-terminal residues of MSPA from WVUTU, when expressed in Escherichia coli, were sufficient to agglutinate erythrocytes (18).

Influenza viruses possess hemagglutinins whose affinities for sialylated receptors differ quantitatively among strains in a pattern functionally constrained by their sialidase activity. Recombinant strains having reduced (36, 37) or deficient (13) sialidase activity and strains passaged in the presence of sialidase inhibitors (7) consequently acquire compensatory mutations in their hemagglutinins that cause decreased affinity for sialylated cellular receptors (16, 17). Mutants revert to the original hemagglutination phenotype following removal of sialidase inhibitors (8), and recombinants expressing hemagglutinins engineered to bind exceptionally tightly to their receptors are strictly dependent on high sialidase activity for the spread of infection (34, 43). Fitness is thus significantly reduced by forces which disturb the quantitative balance between hemagglutinin and sialidase (44), and so lineages which maintain that balance can be expected to predominate among obligate parasites utilizing analogous systems of cytadherence.

The VlhAs and sialidase of M. synoviae are both dominant surface antigens (6). Although numerous residues inferred to experience strong diversifying selection (ω > 3) occur on the exterior surface of the enzyme, strain variation in sialidase activity is not attributable to direct immune-mediated selection (24, 25). We conclude that the VlhAs also experience diversifying selection, such that the species M. synoviae is at its fittest when substantial diversity in the MSPA region of the expressed VlhA is perpetuated among strains (27, 45), and this does seem likely to be the outcome of direct immune-mediated selection. While a shifting quantitative balance with sialic acid-dependent cytadherence may be the proximate selective force constraining sialidase activity in M. synoviae, the implication is that the sialidase gene is a passenger whose diversity is ultimately driven by adaptive host-induced allele switching in vlhA. This could be modeled in vitro by monitoring sialidase activity following antibody-induced vlhA allele switching of isogenic lineages (23). These findings have both diagnostic and therapeutic relevance and may provide new insight into the coevolution of analogous virulence factors in many species of pathogenic bacteria.