A sweeping view of avian mycoplasmas biology drawn from comparative genomic analyses

Abstract

Background

Avian mycoplasmas are small bacteria associated with several pathogenic conditions in many wild and poultry bird species. Extensive genomic data are available for many avian mycoplasmas, yet no comparative studies focusing on this group of mycoplasmas have been undertaken so far.

Results

Here, based on the comparison of forty avian mycoplasma genomes belonging to ten different species, we provide insightful information on the phylogeny, pan/core genome, energetic metabolism, and virulence of these avian pathogens. Analyses disclosed considerable inter- and intra-species genomic variabilities, with genome sizes that can vary by twice as much. Phylogenetic analysis based on concatenated orthologous genes revealed that avian mycoplasmas fell into either Hominis or Pneumoniae groups within the Mollicutes and could split into various clusters. No host co-evolution of avian mycoplasmas can be inferred from the proposed phylogenetic scheme. With 3,237 different gene clusters, the avian mycoplasma group under study proved diverse enough to have an open pan genome. However, a set of 150 gene clusters was found to be shared between all avian mycoplasmas, which is likely encoding essential functions. Comparison of energy metabolism pathways showed that avian mycoplasmas rely on various sources of energy. Superposition between phylogenetic and energy metabolism groups revealed that the glycolytic mycoplasmas belong to two distinct phylogenetic groups (Hominis and Pneumoniae), while all the arginine-utilizing mycoplasmas belong only to Hominis group. This can stand for different evolutionary strategies followed by avian mycoplasmas and further emphasizes the diversity within this group. Virulence determinants survey showed that the involved gene arsenals vary significantly within and between species, and could even be found in species often reported apathogenic. Immunoglobulin-blocking proteins were detected in almost all avian mycoplasmas. Although these systems are not exclusive to this group, they seem to present some particular features making them unique among mycoplasmas.

Conclusion

This comparative genomic study uncovered the significant variable nature of avian mycoplasmas, furthering our knowledge on their biological attributes and evoking new hallmarks.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12864-024-11201-5.

Introduction

Avian mycoplasmas constitute an important polyphyletic group within the Mollicutes class. Many avian mycoplasma species are now recognized, including some of them (Mycoplasma (M.) gallisepticum, M. synoviae, M. meleagridis, M. iowae, and M. imitans) that are generally considered as the most significant pathogens for the poultry industry [1, 2]. Mycoplasmosis clinical manifestations in avians consist predominantly of sinusitis, respiratory disorders, and ocular affections [1, 2]. In contrast to these pathogenic species, some mycoplasmas like M. gallinarum and M. gallinaceum, which are usually accounted among the commensal microbial flora of birds and considered as non-pathogenic, as well as other less known species such as M. anatis, M. columborale, and M. columbinum are often overlooked [2–4].

The recent mycoplasma genomic projects and the several intra- and inter-mycoplasma species whole genome comparisons have significantly contributed to the understanding of the biology and the evolutionary history of these bacteria [5–7]. Like for many other pathogens, studies of avian mycoplasma species have taken advantage of the next-generation sequencing (NGS) power, and many genomes from type and field strains have been sequenced during the last two decades [8–14]. Throughout these genomic reports, many biological and physiological characteristics and numerous key genes related to metabolism and virulence in avian mycoplasmas have been revealed. Some phylogenetic analyses based on different methods (such as concatenated data set or multiple sequence alignment of core genes) have also been proposed in certain studies [9, 14]. Yet, there are still no extensive comparative genomic studies aiming to highlight common and specific traits of these bacteria, which involve multiple phylogenetically distinct and host range-variable species [9].

In this study, we performed a genomic comparison of a selection of 40 strains from 10 different avian mycoplasma species, with variable pathogenicity and/or host specificity. This study intends to provide a broad overview of the biology of the avian mycoplasmas group through comparison and analysis of the publicly available genome sequences.

Results and discussion

Genomic features of avian mycoplasmas

With genome sizes varying from 0.58 to 2.2 Mbp, bacteria belonging to the Mollicutes class are among the smallest self-replicating microorganisms [15]. This is thought to be the result of an extensive genome reduction that they have undergone during their evolution [16, 17]. The genome size of the 40 avian mycoplasmas under consideration in this study fall within this range, with a median size of ~ 0.919 Mbp. The genomes of the four M. meleagridis strains seem to be the smallest among the avian mycoplasmas studied herein. M. meleagridis strains were thus found to have the lowest number of expected coding DNA sequences (CDSs) with a median count of only 528, almost the half compared to the median count of CDSs (954) relative to the three strains of M. iowae, the species having the largest genome (with a median size of 1.23 Mbp). Concerning the intra-species heterogeneity, the most reduced size difference within a complete genome pair of the same species was about 1 Kbp and was observed between M. gallisepticum strains WI01_2001.043–13-2P and NC06_2006.080–5-2P. This difference seems to be insignificant since these two strains were found to have almost the same genetic features. By contrast, the largest intra-species genome size difference, 255 Kbp, was found between the two strains of M. gallinaceum, corresponding to more than 200 CDSs. Intra-species genomic differences could be associated to phenotypic variation such as a switch in virulence potential or in host specificity. However, it should be emphasized that some of the genomes considered here are only drafts assembled obtained from short reads (see Table 1 for details). Therefore, for those genomes, total length and counts of genetic elements might be somewhat misestimated.

Table 1.

General genome features of the avian mycoplasma strains used in the different comparative analyses

Species	Strain	Isolation site	Phylogenetic group	Assembly number	Assembly level	QA (%)b	Genome size (Mb)	GC (%)	Protein-coding	Gene	rRNAc	tRNA	Plasmid
M. anatis	1340	Duck	Hominis	ASM22130v1	44 contigs	95.53	0.928	26.50	775	825	4 (2.1.0)	34	0
	NCTC10156	Duck	Hominis	ASM328506v1	Complete	95.53	0.956	26.50	793	840	6 (2.2.2)	34	0
M. columborale	NCTC10179 a	Pigeon oropharynx	Hominis	57141_A01-3	Complete	99.96	0.966	29.00	749	798	8 (2.3.3)	29	1
M. gallinaceum	B2096 8B	Chicken	Hominis	ASM96576v1	Complete	96.53	0.845	28.50	571	631	5 (1.1.3)	17	0
	NCTC10183 a	Chicken trachea	Hominis	50648_F01-3	Complete	99.05	1.100	28.70	777	837	9 (3.3.3)	30	0
M. synoviae	53	Broiler breeder	Hominis	ASM824v1	Complete	97.18	0.799	28.50	671	734	7 (3.2.2)	34	0
	WVU 1853T	Chicken	Hominis	ASM96976v1	Complete	99.74	0.846	28.50	694	767	7 (3.2.2)	34	0
	MS-H Chicken	Chicken	Hominis	ASM314756v1	Complete	97.10	0.818	28.00	673	740	7 (3.2.2)	33	0
	MS-H Vaccine	Vaccine	Hominis	ASM358608v1	Complete	97.10	0.814	28.00	669	736	7 (3.2.2)	33	0
M. gallinarum	Mgn_IPT	Chicken	Hominis	ASM163723v1	56 contigs	100	0.797	26.00	633	681	4 (2.1.1)	32	0
M. columbinum	NCTC10178 a	Pigeon trachea	Hominis	57572_G01-3	Complete	96.55	0.776	27.00	602	647	7 (3.2.2)	32	0
	SF7	Oral and pharyngeal swabs of an apparently healthy piegon	Hominis	ASM22299v1	15 contigs	93.10	0.748	27.00	596	645	5 (3.1.1)	32	0
M. meleagridis	ATCC 25294 (17529)	Turkey	Hominis	ASM96962v1	22 contigs	98.10	0.634	26.00	520	568	4 (2.1.1)	33	0
	MM_26B8_IPT	Chicken	Hominis	ASM163966v1	32 contigs	99.05	0.658	26.50	545	600	4 (2.1.1)	33	0
	NCTC10153 a	Turkey	Hominis	57612_E01-3	Complete	99.05	0.644	26.00	519	562	6 (2.2.2)	33	0
	IZSVE/2944/9/2011	Turkey	Hominis	Mmel2011	9 scaffolds	99.05	0.647	26.00	530	574	2 (2.0.0)	34	0
M. iowae	695	Turkey	Pneumoniae	ASM22735v1	70 contings	95.99	1.200	24.50	933	982	3 (1.1.1)	29	0
	NCTC10185 a	Air Sac of a pipped embryo	Pneumoniae	51670_F01-3	Complete	97.90	1.300	24.50	993	1034	3 (1.1.1)	29	0
	DK-CPA	Tracheae air sacs and pericardial regions from turkey embryos	Pneumoniae	MycoIowaeDkcpa1.0	89 contings	96.47	1.200	24.50	937	983	3 (1.1.1)	29	0
M. imitans	ATCC 51306 a	Turbinate of duck	Pneumoniae	ASM51830v1	24 scaffolds	98.91	0.922	30.50	700	772	2 (2.0.0)	31	0
M. gallisepticum	Field strain Rlow	Chicken	Pneumoniae	ASM9258v1	Complete	98.86	1.000	31.50	750	818	6 (2.2.2)	32	0
	Field strain Rhigh	Chicken (attenuated strain)	Pneumoniae	ASM2536v1	Complete	97.90	1.000	31.50	738	810	6 (2.2.2)	32	0
	Field strain F	Chicken (vaccine strain)	Pneumoniae	ASM2538v1	Complete	99.02	0.977	31.50	748	809	6 (2.2.2)	32	0
	S6	Chicken	Pneumoniae	ASM21154v6	Complete	96.12	0.985	31.50	759	828	6 (2.2.2)	33	0
	VA94_7994-1-7P a	House finch	Pneumoniae	ASM28667v1	Complete	96.54	0.964	31.50	732	789	6 (2.2.2)	32	0
	NC95_13295-2-2P	House finch	Pneumoniae	ASM28669v1	Complete	96.54	0.954	31.50	721	777	6 (2.2.2)	32	0
	NC96_1596-4-2P	House finch	Pneumoniae	ASM28671v1	Complete	96.54	0.986	31.50	730	791	6 (2.2.2)	32	0
	NY01_2001.047–5-1P	House finch	Pneumoniae	ASM28673v1	Complete	96.54	0.965	31.50	723	784	6 (2.2.2)	32	0
	WI01_2001.043–13-2P	House finch	Pneumoniae	ASM28675v1	Complete	96.38	0.939	31.50	708	767	6 (2.2.2)	32	0
	NC06_2006.080–5-2P	House finch	Pneumoniae	ASM28677v1	Complete	96.31	0.938	31.50	708	768	6 (2.2.2)	32	0
	CA06_2006.052–5-2P	House finch	Pneumoniae	ASM28679v1	Complete	95.98	0.976	31.50	724	786	6 (2.2.2)	32	0
	NC08_2008.031–4-3P	House finch	Pneumoniae	ASM28681v1	Complete	96.31	0.926	31.50	703	762	6 (2.2.2)	32	0
	K6356	Tracheae of domestic chicken	Pneumoniae	ASM167649v1	17 contigs	96.37	0.957	31.50	632	798	4 (2.1.1)	32	0
	K6372	Tracheae of domestic chicken	Pneumoniae	ASM167650v1	69 contigs	96.32	0.948	31.50	722	832	1 (1.0.0)	32	0
	K5322C	Tracheae of domestic chicken	Pneumoniae	ASM167651v1	90 contigs	96.67	0.943	31.50	718	832	2 (2.0.0)	31	0
	K6112B	Tracheae of domestic chicken	Pneumoniae	ASM167652v1	94 contigs	96.67	0.928	31.50	714	808	4 (2.1.1)	32	0
	K2966	Tracheae of domestic chicken	Pneumoniae	ASM167657v1	71 contigs	95.93	0.958	31.50	723	831	2 (2.0.0)	32	0
	K6222B	Tracheae of domestic chicken	Pneumoniae	ASM168363v1	82 contigs	96.67	0.941	31.50	719	816	2 (2.0.0)	32	0
	K6216D	Tracheae of domestic chicken	Pneumoniae	ASM168367v1	92 contigs	96.30	0.943	31.50	724	827	2 (2.0.0)	32	0
	K6208B	Tracheae of domestic chicken	Pneumoniae	ASM170574v1	95 contigs	94.07	0.948	31.50	719	830	2 (2.0.0)	32	0

All data are gathered from NCBI database

^a Sequences of strains with this symbol are designed as reference genomes by NCBI. Underlined strains are those sequenced by our team

^b Quality analysis: Quality of genomes was assessed in NCBI using the automated method “CheckM” (v1.2.3) for all the strains, except the strain M. gallinaceum B2096 8B for which the quality was evaluated using another technique (ANI: Average Nucleotide Identity)

^c Only complete rRNAs are considered. The absence of some rRNAs in some strains is probably not reflecting the reality and could be explained by assembly or annotation errors. Numbers in brackets refer to 5S rRNA, 16S rRNA, and 23S rRNA genes, respectively

Like in most of the Mollicutes species [15], the G + C content in the genomes of the avian mycoplasmas was low, ranging from 24.50% (M. iowae strains) to 31.50% (M. gallisepticum strains). GC% were either constent or slighlity varing between strains belonging to a same avian mycoplasma species (Table 1). Although, GC% observed in M. iowae strain 695 is relatively low, it still falls within the range of the genus Mycoplasma (23%—40%) [15].

A reduced but sufficient set of genes encoding transfer RNAs (tRNAs), from 29 to 34, corresponding to the full range of the 20 amino acids (aa) was found in all avian mycoplasmas, expect for M. gallinaceum strain B2096 8B in which only a total of 17 tRNAs was defined (Table 1). This uncommon reduced tRNA count in this mycoplasma strain is most probably due to an artefact caused by misassembly. This was reported in the genome notes relative to the genome assembly of this strain in the NCBI database. Overall, the number of tRNA genes in avian mycoplasmas, like in the other Mollicutes class members, is kept to a minimum and that the level of redundancy of these genes is low with few, if any, gene duplication [15, 18]. This finding contrasts with the high redundancy of tRNA genes observed in other bacteria with larger genomes such as Escherichia (E.) coli, Bacillus subtilis, and Haemophilus influenzae with 86, 86, and 54 tRNA genes, respectively [19–21].

Regarding ribosomal RNAs (rRNAs), all the 40 avian mycoplasma strains were found to harbour these genes, but with different set numbers (as shown in Table 1). It is worth reminding that only complete rRNA genes are taken into account. Thus, the absence of some rRNAs in some strains is probably not reflecting the reality and could be explained by assembly or annotation errors. It has previously been reported that no more than one or two copies of the rRNA genes were usually detected in the members of the Mollicutes class [22]. This was not the case for M. gallinaceum species. Actually, the sequenced genome of M. gallinaceum strain NCTC10183 (released on 2019) was found to harbour three copies of each rRNA gene (5S, 16S, and 23S), a feature which might explain its fast-growing nature [23]. By analogy to what has been previously reported in E. coli, the multiple rRNAs observed in M. gallinaceum may possibly confer to the bacterium a selective advantage permitting a capacity for higher growth rates particularly in environments with fluctuating resources [24]. Interestingly, a recent study showed that the capacity for rapid growth is not the sole purpose of rRNA operon multiplicity in E. coli. It was proved that the high number of these operons plays also a role in guaranteeing genome stability by reducing DNA replication blockages [25]. This scenario can be hypothesized for this M. gallinaceum strain, but with no certainty as no data about organization of rRNA genes in operons in this mycoplasma species is available.

A thorough examination of M. gallinaceum strain NCTC10183 genome showed that the increased copy number of rRNA genes has likely arisen from a duplication event. Indeed, the two pairs of 23S rRNA—16S rRNA gene sequences (NCTC10183_00253—NCTC10183_00254) and (NCTC10183_00255—NCTC10183_00256) were nearly identical (more than 98% identities) and were placed next to each other in the genome of M. gallinaceum strain NCTC10183. A third pair of 16S-23S rRNA-encoding genes (NCTC10183_00523—NCTC10183_00524) with different orientation was also detected in the genome of this mycoplasma strain, but far distant from the two first pairs. Likewise, three 5S rRNA gene copies (NCTC10183_00480, NCTC10183_00481, and NCTC10183_00443) whose sequences proved nearly identical (99% identity) were detected. The two first copies were juxtaposed but the third was detected in a distant position in the genome of M. gallinaceum strain NCTC10183. The remote location of rRNAs in this strain is not unusual among Mycoplasma spp. Indeed, these ribosomal genes were previously found to be either linked under the typical prokaryotic pattern 5′16S-23S-5S 3’ in some species or unlinked with large spacers between them in others [24]. Moreover, duplication events of these sets of rRNA genes could occur fortuitously during DNA replication and/or repair, and have previously been described for genes other than those encoding rRNAs. Indeed, the use of operon system in avian mycoplasma species has also been observed for many other genes, like for instance, the haemagglutinin-encoding genes [26]. Other striking examples are the genes gapA (formerly mgc1) and mgc2 involved in cytadherence in M. gallisepticum [27, 28] and the couple of genes encoding for the Mm19 AlwI nuclease and the methyltransferase (forming together a type II restriction-modification system) in M. meleagridis [29]. The explanation of this frequent use of operon system in mycoplasmas might be the high gene density usually observed in their genomes, which potentially reduces the number of regulatory factors needed to control gene transcription [15].

The existence of extrachromosomal DNAs (plasmids) has been reported only in a few Mollicutes species isolated from different hosts such as humans [30], arthropods [31], or caprine [32], but never from birds. According to its genome sequence, the pigeon mycoplasma species, M. columborale, seems to be the first described avian mycoplasma to carry a plasmid. This extrachromosomal DNA found in M. columborale NCTC10179 strain has a size of about 17 Kbp with low GC content (26.5%). Seventeen predicted CDSs were identified. Thirteen out of them were annotated as uncharacterized proteins and functions were assigned to only four CDSs; Anhydro-N-acetylmuramic acid kinase (NCTC10179_00787), Transposase (NCTC10179_00790), Integrase-recombinase protein (NCTC10179_00797), and Oligoendopeptidase F (NCTC10179_00803). Nevertheless, no typical plasmid features such as the existence of replication protein were detected. Also, no similarity was found between the nucleotide sequence of this avian mycoplasma plasmid with any previously reported plasmid in other mycoplasmas or bacteria. A significant matching between the DNA sequences of intra- and extra-chromosomal genetic materials was revealed by BLAST. In fact, the whole plasmid sequence was found in the chromosome of M. columborale, but in two separate DNA blocks. This could put into doubt the real existence of this plasmid since it is known that the extra- and intra-chromosomic materials should normally be different. Further in silico analyses and supplementary experimental techniques are needed to definitely establish the nature of this genetic material.

Phylogenetic analysis and host specificity

Taxonomically, mycoplasmas belong to the class Mollicutes [15]. This class is divided into five phylogenetic groups: Hominis, Pneumoniae, Spiroplasma, Anaeroplasma, and a group known to contain the only species Asteroleplasma anaerobium [33]. Genome-based phylogenies of Mollicutes have been proposed as soon as sufficient genomics data became available [17, 34]. Here, we have resolved an accurate phylogenetic tree by concatenating the alignments of 47 core proteins involved in translation and occurring in all Mollicutes species (Fig. 1, Additional file 1). In this tree, generated using the maximum likelihood method, we have included the 40 avian mycoplasma strains investigated in this study, as well as some representative Mollicutes species of each phylogenetic group and the Gram positive bacterium Bacillus subtilis as an outgroup. The tree showed that the avian mycoplasma species, analysed in this work, belong either to the Hominis or Pneumoniae phylogenetic groups. Apart from M. iowae, M. gallisepticum, and M. imitans that belong to the Pneumoniae group, all the other avian mycoplasmas fell into the Hominis group. Within each group, two clusters could be defined. In the Hominis group, the first cluster is composed of M. anatis, M. columborale, M. gallinaceum, and M. synoviae species, and the second is made of M. gallinarum, M. columbinum, and M. meleagridis species. In the Pneumoniae group, M. iowae species was placed in a cluster that is different from the one comprising M. imitans and M. gallisepticum species. This phylogenetic segregation in groups and clusters seems not to be linked to host specificity. Indeed, many avian mycoplasma species infecting different bird hosts (duck, chicken, pigeon …) were assembled together, while species sharing the same host were found to be distantly placed in the phylogenetic tree. For example, M. gallisepticum and M. imitans were found to have a very tight phylogenetic link despite of having different host specificities (chicken/house finch and duck, respectively). By contrast, M. columborale and M. columbinum, which are both mycoplasmas of pigeons, belong to two different clusters in the Hominis group.

Fig. 1.

Phylogenetic tree of Mollicutes. The phylogenetic tree was generated using the maximum likelihood method from the concatenated multiple sequence alignments of selected 47 orthologous protein involved in translation. Main phylogenetic groups are indicated. S, Spiroplasma, AAP, Acholeplasma/Phytoplasma. Clusters are delimited by brackets. Bird mycoplasmas are highlighted in blue. Bacillus subtilis was used as an outgroup. Bootstrap statistical values are indicated on branches. Triangles and circles indicate fermentative and non-fermentative avian mycoplasma species, respectively. Species capable of using both metabolic pathways are indicated by rhombuses

Overall, the present phylogenetic analysis further refines and confirms the results of earlier studies, especially those based on 16S rRNA. It provides a more valid, robust, and accurate estimation of the evolutionary relationship between the avian mycoplasmas with inclusion of new sequenced strains (notably the three strains M. meleagridis ATCC 25294, MM_26B8_IPT, and Mgn_IPT sequenced by our team). Concerning host specificity, we found no host co-evolution of avian mycoplasma species within the Mollicutes class. Actually, host specificity is rather explained by functional differentiations in surface-exposed protein(s) that interact with the host receptor(s). It was suggested that mycoplasmas are probably acquired by their hosts through an intimate contact followed by a transfer of material between mucosal surfaces. Mycoplasmas are often reported to be host-specific. However, exceptions were reported in some species including bovine and caprine mycoplasmas [35]. Host crossing was reported in avian mycoplasmas as well. The isolation of M. meleagridis, a turkey pathogen, from chicken is a striking example [13]. More impressive are the results of a meta-analysis study reporting the presence of M. gallisepticum in 56 species of bird belonging to 11 different orders including Galliformes, Falconiformes, and Passeriformes [36]. Overall, with the tremendous omics data that have become available during the last years, there is clearly a better understanding about the mechanisms and strategies used by mycoplasmas to survive hostile environments and adapt to new hosts [37].

Core essential genes and pan genome of avian mycoplasmas

Given their basic metabolism machinery and their low genomic redundancy, mycoplasmas represent a good model to study and define the minimal set of genes needed to sustain bacterial life [38, 39]. In the present study, our analyses of pan and core genomes focused on three different sets. Firstly, we have considered the totality of the 40 mycoplasma strains studied herein as a single entity representing the avian mycoplasmas group. In the second and the third analyses, we have partitioned the avian mycoplasmas according to their phylogenic placement within the Mollicutes class in the Hominis (16 avian mycoplasma strains) and Pneumoniae (24 avian mycoplasma strains) groups, respectively.

The number of CDSs per genome within the various 10 avian mycoplasma species is ranging from 519 (in M. meleagridis strain NCTC10153) to 993 (in M. iowae strain NCTC10185) with an average of ~ 707 and a total of 28,295 predicted proteins. Based on sequence similarity, all these putative proteins were classified into 3,387 clusters (Fig. 2, Panel A). Only 150 clusters were found to be shared between all the 40 genomes, representing the core set of avian mycoplasmas group (Fig. 2, Panel Ac). These intersected clusters (corresponding to less than 5% of the total number of clusters) mainly encode proteins that are essential for the survival of avian mycoplasmas. Typically, the core genome of mycoplasmas contains genes responsible for their major phenotypic traits and encoding for the proteins involved in essential cellular functions such as ribosomal synthesis, DNA replication, transcription and translation, energy generation, and some metabolism machineries [7].

Fig. 2.

Pan genomes and core genomes of avian mycoplasmas. A Set of all avian mycoplasmas studied here (40 strains). B Set of avian mycoplasmas belonging to Pneumoniae group (24 strains). C Set of avian mycoplasmas belonging to Hominis group (16 strains). (a) Gene number estimation for the pan genome (growing curve). The number of specific genes is plotted as a function of the number of strains sequentially added. (b) Gene number estimation for the core genome (downward curve). The number of shared genes is plotted as a function of the number of strains sequentially added. (c) Prevalence of the different gene clusters across the avian mycoplasma strains

Moreover, when we applied the same analysis on avian mycoplasmas split into two groups according to their phylogenetic positioning within the Mollicutes class, we found that the core sets of homologous gene clusters were greater. The core genome of the avian mycoplasmas belonging to the Hominis group was found to be composed of 250 clusters (~ 12% of the total of 2,116 homologous gene clusters) (Fig. 2, Panel Cc). As for the Pneumoniae group, the number of conserved gene clusters was the highest. It was about 300, representing 17% of the total of 1764 homologous gene clusters (Fig. 2, Panel Bc). Hence, it seems that the avian mycoplasmas belonging to the Pneumoniae group are less variable than those belonging to the Hominis group. This could be explained by the fact that the 24 mycoplasma strains affiliated to this group belong to only 3 species; M. gallisepticum (with a majority of 20 strains, all isolated from limited isolation sites), M. iowae (with 3 strains), and M. imitans (with one strain). By contrast, the 16 strains of the Hominis group belong to 7 different avian mycoplasma species (M. meleagridis, M. gallinarum, M. gallinaceum, M. anatis, M. synoviae, M. columborale, and M. columbinum) having various hosts (chicken, turkey, duck, and pigeon). This difference observed in core size is not without precedent. Indeed, it was previously reported that the core set genes count and composition could be contingent upon the phylogeny and the number of included genomes. More precisely, it was proposed that the core size of an organism group is inversely proportional to the number of genomes and the phylogenetic diversity among this group [40]. This was exactly observed in the three tested mycoplasma sets. The downward curves represented in the Fig. 2 (Panels Ab, Bb, and Cb) correspond to the decline of the core size as a function of the number of mycoplasma genomes taken into account.

Soft cores, which represent genes found in ≥ 95% of accessions, were about 1/3 of core genomes in the group of all avian mycoplasmas together (~ 50 vs ~ 150) as well as in the avian mycoplasmas belonging to the Hominis group (~ 80 vs ~ 250). However, for the avian mycoplasmas belonging to the Pneumoniae group, the soft core appeared to be too small comparing to the core genome (~ 25 vs ~ 300). This big difference may be explained by the fact that the genome sequence of several strains belonging to this group is incomplete and so the number of genes is also incomplete. As for the remaining gene clusters (cloud and shell pan genomes) constituting the accessory genome of each of the three avian mycoplasma sets, most of them (between 70 and 80%) were either unique to a single strain or detected in two or three strains. This is in concordance with the fact that the number of unique or specific genes is vast [41]. In contrast to the core genome, the pan genome increases with the number of the analyzed strains. The growing curve (Fig. 2, Panel Aa) suggests that the avian mycoplasma group taken as a whole has an open pan genome. Thus, one can argue that the gene diversity within this group of mycoplasmas is far from being fully explored, since the number of newly discovered gene clusters is permanently increasing as a function of newly sequenced avian mycoplasma genomes. This open-shaped pan genome was also observed for the avian mycoplasmas belonging to the Hominis group and those belonging to the Pneumoniae group (Fig. 2, Panels Ba and Ca). This suggests that even in a restricted phylogenetic group, an important diversification of gene sets could occur. Notably, this observation was more pronounced in the Pneumoniae group despite being almost totally composed of strains belonging to only one species (M. gallisepticum).

Actually, the infinite or finite nature of the pan genome of a given bacterial species is mainly bound to its capacity to acquire and use exogenous DNA, a fact that is intimately linked to its lifestyle [41, 42]. In a previous study including 20 different mycoplasma species genomes, the huge pan-genome size, composed of more than 8,000 genes, was thought to be a reflection of the diversity of their lifestyles in different ecological niches [7].

With 3,237 different gene clusters found dispersed in the different 40 avian mycoplasma strains investigated herein, the avian mycoplasma group is proved to be versatile and diverse enough to have an open pan genome. Yet, the restricted number of only 150 gene clusters shared by all the 40 avian mycoplasma genomes, which thus code for vital functions, goes along the idea considering mycoplasmas as minimal organisms.

Energy-yielding pathways

Because of their reduced genome size, the Mollicutes are known to have a limited metabolic potential. They are serving as models for describing the minimal metabolism necessary to sustain an independent life [43, 44]. Due to the absence of tricarboxylic acid (TCA) cycle and functional respiratory chain, the Mollicutes need to rely on other pathways to generate energy. Energy source has been a main criterion for classifying and characterizing Mollicutes [45]. Some species are able to metabolize sugar, others can breakdown arginine, and some others are capable of energy production by urea hydrolysis or by oxidation of organic acids such as lactate and pyruvate [46–48]. Generally, the Mollicutes are divided into two broad groups; fermentative and non-fermentative based on their ability to metabolize carbohydrates [15, 49]. Here, we partitioned the avian mycoplasma species into three groups according to their metabolic characteristics as described in the Bergey’s manual [50]. The first energy metabolism group comprised the 6 glycolytic species which are M. anatis, M. columborale, M. gallinaceum, M. imitans, M. synoviae, and M. gallisepticum. The second group included the 3 species capable of hydrolysing the arginine; M. gallinarum, M. columbinum, and M. meleagridis. The third and last group contained only one species, M. iowae that was classified as a species capable of utilizing both glucose and arginine as energy sources [50, 51]. This segregation of avian mycoplasmas into groups based on their energy metabolism was overlayed with the phylogenetic clusters to determine if any correlation exists between metabolism and phylogeny for this set of mycoplasmas (See geometric symbols in Fig. 1). The superposition analysis showed that the glycolytic mycoplasma species belong to two phylogenetic groups (Hominis group, cluster 1 and Pneumoniae group, cluster 2). In contrast, all the arginine-utilizing mycoplasma species were found to belong to the same cluster within the same group (cluster 2 of Hominis group). This probably suggests that the phylogenetic relatedness between the arginine-utilizing species is more tight comparing to sugar-utilizing species.

Genes encoding proteins involved in energy generation were identified using targeted computer-aided searches in genome sequences of the above-listed avian mycoplasmas (Fig. 3). Details about genes loci are shown in Additional files 2–5 for the 1st group, Additional file 6 for the 2nd group, and Additional file 7 for the 3rd group.

Fig. 3.

Energetic pathways comparison in avian mycoplasmas group. Numbers from 1 to 23 correspond to genes encoding for proteins involved in energetic pathways in avian mycoplasmas: ADI (Arginine dihydrolase) pathway. 1: arcA (Arginine deiminase); 2: arcB (Ornithine carbamoyltransferase); 3: arcC (Carbamate kinase). EMP (Embden-meyerhof-parnas) pathway. 4: pgi (Glucose-6-phosphate isomerase); 5: pfk (6-phosphofructokinase); 6: fba (Fructose biphosphate aldolase); 7: tpiA (Triose-phosphate isomerase); 8: gap (Glyceraldehyde 3-phosphate dehydrogenase); 9: pgk (3-phosphoglycerate kinase); 10: pgm (Phosphoglycerate mutase); 11: eno (Enolase); 12: pyk (Pyruvate kinase). PDH (Pyruvate dehydrogenase) complex. 13: pdhA (Pyruvate dehydrogenase E1 α subunit); 14: pdhB (Pyruvate dehydrogenase E1 β subunit); 15: pdhC (Dihydrolipoamide acetyltransferase E2); 16: pdhD (Dihydrolipoamide dehydrogenase E3). PPP (Pentose phosphate pathway). 17: g6pd (Glucose-6-phosphate dehydrogenase); 18: pgls (6-phosphogluconolactonase); 19: pgd (6-phosphogluconate dehydrogenase); 20: ripB (Ribose-5-phosphate isomerase); 21: rpe (Ribulose 5-phosphate 3-epimerase); 22: tkt (Transketolase); 23: Taldo1 (Transaldolase). Abbreviations mentioned in the left refer to the 40 avian mycoplasma strains studied here. Strains were delimited into brackets according to their metabolic pathway as follows: 1st group (Glucose-utilizing avian mycoplasmas). Ma1 and Ma2: Mycoplasma anatis strains 1340 and NCTC10156, respectively. Mco: Mycoplasma columborale strain NCTC10179. Mgc1 and Mgc2: Mycoplasma gallinaceum strains B2096 8B and NCTC10183, respectively. Mi: Mycoplasma imitans strain 51306. Ms1-Ms4: Mycoplasma synoviae strains 53, WVU 1853^T, MS-H (chicken), and MS-H (vaccine), respectively. MG1-MG20: Mycoplasma gallisepticum strains Rlow_, Rhigh_, F, S6, VA94_7994-1-7P, NC95_13295-2-2P, NC96_1596-4-2P, NY01_2001.047–5-1P, WI01_2001.043–13-2P, NC06_2006.080–5-2P, CA06_2006.052–5-2P, NC08_2008.031–4-3P, K6356, K6372, K5322C, K6112B, K2966, K6222B, K6216D, and K6208B, respectively. 2nd group (Arginine-utilizing avian mycoplasmas). Mgn: Mycoplasma gallinarum strain Mgn_IPT. Mci1 and Mci2: Mycoplasma columbinum strains NCTC10178 and SF7, respectively. Mm1-Mm4: Mycoplasma meleagridis strains 25294, MM_26B8_IPT, NCTC10153, and IZSVE/2944/9/2011, respectively. 3rd group (Glucose and arginine utilizing-avian mycoplasmas). Mio1-Mio3: Mycoplasma iowae strains 695, NCTC10185, and DK-CPA, respectively. Strains with complete genome are printed in bold type. Colored cells refer to the existence of the corresponding gene involved in energetic pathways. The absence of a gene is noticed by a blank or grey-colored cells for strains with complete and draft genomes, respectively

Genome screening of the 30 avian mycoplasma strains belonging to the fermentative group showed that, aside from M. gallinaceum strain B2096 8B, they all carry the whole set of enzymes constituting the Embden-Meyerhof-Parnas (EMP) pathway, which confirms their ability to perform glycolytic metabolism. In contrast with M. gallinaceum strain NCTC10183, strain B2096 8B was found to lack the first enzyme (glucose-6-phosphate isomerase) of the glycolysis process (Additional file 2). Actually, the absence of such an important enzyme could be due to a sequencing or an annotation error since the ability of M. gallinaceum species to ferment glucose has been experimentally proven far before [51]. In each of the fermentative strains, all the detected glycolytic genes were found to be dispersed in their genomes. This suggests that the expression of these genes is probably executed and regulated independently. The EMP pathway consists mainly in glucose catabolism and production of energy in the form of two adenosine 5'-triphosphate (ATP) molecules. The final product of this pathway is the pyruvate. This organic acid could be metabolized to acetyl coenzyme A (acetyl-CoA) by the pyruvate dehydrogenase (PDH) complex [49, 52]. The genes pdhA, pdhB, pdhC, and pdhD encoding the enzymes (pyruvate dehydrogenase E1 alpha and beta subunits, dihydrolipoamide acetyltransferase E2, and dihydrolipoamide dehydrogenase E3, respectively) of this complex were detected in all the glycolytic avian mycoplasmas. Apart these latters, avian mycoplasmas belonging to the second and third energy metabolism groups, except for M. meleagridis strains, were also found to harbour these four genes of PDH complex. These genes were found to be clustered in a single region practically in almost all the fermentative species. This could be of some regulatory significance and goes with their arrangement in one operon. This organization is not specific to avian mycoplasmas since it has been similarly reported before for the PDH complex in other Mollicutes species not isolated from birds like M. capricolum [53] and Acholeplasma laidlawii [54] and many more. Other different arrangements of the PDH complex genes were described before within the Mollicutes class. In M. hyopneumoniae for instance, the genes quadruplet of the PDH complex was found to be organized in two distinct operons [55].

Moreover, it is worth reminding that despite the production of pyruvate at the end of the glycolysis process, the EMP pathway is not linked to the Krebs cycle in these avian mycoplasmas, alike all the other mycoplasmas, since the enzymatic machinery of this cycle is completely deficient [15, 49]. The pentose phosphate pathway (PPP), which is a process of glucose turnover (principally producing pentoses that are necessary for nucleotides metabolism), was also found to be incomplete in all the avian mycoplasmas mentioned in this study. This truncated pathway may be a result of the genome reduction that mycoplasmas have been subjected to as it has also been reported for most the other mollicutes with the exception of Acholeplasma species [45].

Surprisingly, we have observed the presence of the arcA gene in M. imitans and M. gallisepticum strains. This gene encodes for the arginine deiminase; an enzyme belonging to the arginine dihydrolase (ADI) pathway normally found in arginine utilizing-mycoplasma species. This gene detection was noticeable since these species are fermentative. Most strickling is the presence of this gene in more than one copy in some strains (Additional files 2–5). Some versions of this gene were complete, others were either truncated or just considered as pseudogenes. The in silico detection of an approved arcA gene in glucose using-mycoplasmas such as M. imitans and M. gallisepticum could theoretically implies the existence of an arginine deiminase in these species. Indeed, the outputs of a metabolomic analysis including M. gallisepticum species seems to confirm this assumption. The arginine deiminase, which is the central enzyme of the ADI pathway, was not only identified in this avian mycoplasma species, but also proved to be functional as a high concentration of citruline, an intermediate in the ADI pathway, has previously been detected [56]. Vanyushkina et al. reminded in this study that an adaptation to acidic environment stress conditions is necessary for the avian pathogen M. gallisepticum to succeed its host infection, and they experimentally proved that this species can survive at low pH values (less than 5) [56]. Yet, the non-detection of any mark of arcB and arcC genes coding for the two other enzymes of the ADI pathway comes showing the incompleteness of this pathway and thus the uncertainty about its utilization as an alternative energy source in glycolytic mycoplasma species. The presence of genes coding for some enzymatic components of the ADI pathway in M. imitans and M. gallisepticum is unique among the fermentative avian mycoplasmas group, but not among all the fermentative mollicutes species. As a matter of fact, the same case has been reported before for the human glycolytic mycoplasma species M. pneumoniae. In such a study, a bacteria transformation followed by enzymatic tests showed that M. pneumoniae transformants possessing the three functional ADI enzymes were unable to grow on arginine as a sole energy source [45]. Wodke et al. have also reported that M. pneumoniae can use arginine to produce ATP, but due to its insignificant contribution to energy production, this was not believed to be an actual alternative carbon source [57]. Likewise, results from Yus et al. came to the same conclusion, but in vivo [58]. The truncated version of the ADI pathway in this human pathogen was explained by the deletion of significant regions in the proteins ArcA and ArcB, which has led to the lose of their enzymatic activities [45]. All of the aforementioned could be imagined for M. imitans and M. gallisepticum species. From a phylogenetical point of view, these two avian mycoplasma species are very close to M. pneumoniae. All three of them are belonging to the Pneumoniae group within the Mollicutes class. This could support the hypothesis that these species have underwent the same evolutionary scenario with regards to the gain and loss of the ADI pathway genes.

By contrast, the ADI pathway was integrally identified in all the 7 non-fermentative avian mycoplasma strains belonging to M. gallinarum, M. columbinum, and M. meleagridis species (second energy metabolism group) (Additional file 6). This is consistent with the ability of these species to perform arginolysis to generate energy. Through this pathway, arginine hydrolysis results in the production of ornithine, ATP, CO₂, and ammonia, which raises the pH of the culture medium [49]. The question whether arginine degradation can be sufficient or not to supply the required energy to the cells of non-fermentative Mollicutes has been raised before, and the role of the ADI pathway as unique source of energy has been previously interrogated [59, 60]. It has been shown that the growth of non-fermentative ADI-positive mycoplasmas was remarkably stimulated when L-arginine is added to the medium [61, 62]. This result proved that arginine can, by itself, satisfy the energy requirements of non-fermentative mycoplasmas. Added to this, it has also been demonstrated that no ATP energy is needed for arginine uptake into the mycoplasma cells [63]. Taken together, these findings confirm that the ADI pathway is energetically advantageous for the non-fermentative mycoplasmas. The genes arcA, arcB, and arcC encoding for the three enzymes (arginine deiminase, ornithine carbamoyltransferase, and carbamate kinase, respectively) involved in the ADI pathway were found contiguous in M. gallinarum, M. columbinum, and M. meleagridis species (Fig. 4, Panel A). This gene synteny is slightly different from what was described before for one of the best studied arginine-utilizing mycoplasma species, the human pathogen M. hominis (PG21 strain) [64]. Yet, it could not be considered as specific to avian mycoplasmas within the Mollicutes class since a quite similar organization of the complete set of the ADI pathway genes has previously been reported in another human pathogen (M. penetrans HF-2 strain) [65]. The close arrangement of arginine metabolism genes indicates that they are seemingly organized in an operon, which probably facilitates the coordinate regulation of the enzymes they encode for. In vicinity of the arcABC locus, a gene encoding for a permease belonging to the APC (amino acid-polyamine-organocation) family was identified in M. gallinarum Mgn_IPT strain (MGALLINA_RS00420), in both M. columbinum strains (EXC37_RS01340 and MCSF7_RS03335), and in the four strains of M. meleagridis (MMELEA_RS01145, MM26B8_RS01870, EXC33_RS01580, and ASB56_RS01900) (Fig. 4, Panel A). BLASTP followed by PSIPRED and MEMSAT-SVM analyses showed that the permease protein harbours a putative conserved domain (AA_permease_2: pfam13520) carrying 12 transmembrane segments. The transmembrane nature of this protein as well as its close localization to arcABC cluster suggest its association to the uptake and the transportation of arginine into the cell [64].

Fig. 4.

Layout of CDSs of the arcABC operon in arginine-utilizing avian mycoplasma species (A) and in arginine/glucose-utilizing avian mycoplasma species (B). Gray boxes refer to CDSs around the arcABC operon. The other boxes were colored according to the color code of the genes arcA, arcB, arcC, arcD, and Permease as illustrated in the figure. Data are extracted from Molligen and NCBI databases. Double slash indicates discontinuation of genomic sequences

In spite of being nonglycolytic species, the arginine-using species M. meleagridis and M. columbinum were found to contain many genes coding for enzymes implicated in the carbohydrates metabolism. However, the EMP pathway was truncated in both of them (Fig. 3). In fact, the pgiB, pfk, and fba genes encoding the three first enzymes of the glycolysis process (glucose-6-phosphate isomerase, 6-phosphofructokinase, and fructose biphosphate aldolase, respectively) are missing in M. meleagridis strains. As for M. columbinum strains, only pfk and fba genes were lacking. In contrast, the EMP pathway was found to be complete in M. gallinarum (Fig. 3 and Additional file 6). According to these predictions, it would appear that arginine metabolism is the sole energy-generating mechanism in M. meleagridis and M. columbinum, but not in M. gallinarum, which seems able to produce ATP by glycolysis too. This finding was unexpected since M. gallinarum has been classified as an arginine nonfermenting species [50]. In addition, according to the best of our knowledge, the ability to use glucose was never challenged before for any specific M. gallinarum strain. Thus, it would be of great interest to explore the EMP pathway in M. gallinarum to evaluate the weight that it could have in the energy metabolism and to verify that the genes of glycolytic pathway are effectively expressed. Rigorous examination and monitoring of growth behaviour of this species in different culture media compositions would be very helpful to get with conclusive idea regarding energy generation in M. gallinarum species. A comparison with the energy genetic background of M. iowae should also be conductive as this avian mycoplasma species is the only one for which the ability of utilizing arginine and of fermenting glucose aerobically and anaerobically has been experimentally proven before [51, 59]. This double capacity was genetically sustained by the simultaneous detection of the full set of enzymes of ADI and EMP energy-yielding pathways in M. iowae strains 695, NCTC10185, and DK-CPA (Additional file 7). By including these three strains of M. iowae (constituting the third energy metabolism group) into the synteny analysis of the arcABC operon described above for the arginine-utilizing mycoplasmas, a similarity in genes organization has been obviously noticed (Fig. 4, Panel B). This striking similarity may be due to the conservation of this set of genes since it was found almost identical in mycoplasma species from two different phylogenetic groups (Hominis and Pneumoniae). Alike mycoplasmas of the second energy metabolism group, M. iowae strains were also found to contain APC permeases harbouring the AA_permease_2 putative conserved domain and carrying 12 transmembrane helices (as revealed by MEMSAT-SVM analysis in PSIPRED online workbench). Nevertheless, these permeases were found to be in greater numbers (four in each strain) and to be scattered in the genomes (not contiguous to the arcABC operon like in the arginine utilizing-mycoplasmas) (Fig. 4, Panel B). BLASTP analysis showed that two of these permeases have homologs in other avian mycoplasmas, notably M. imitans and M. gallisepticum but with low similarity percentages (30%—40%). However, the two other permeases were exclusive to M. iowae species; one among the avian mycoplasmas group and the other among the whole Mollicutes class. These findings may reflect a kind of particularity and development in the aa transport system of this species. This was further strengthened by the detection of other loci (GUU_RS03580, EXC57_RS04235, and P271_RS00165) in M. iowae strains 695, NCTC 10185, and DK-CPA, respectively encoding for a YfcC protein which is a member of the basic aa antiporter (ArcD) family (Fig. 4, Panel B). This antiporter, also called exchanger and labelled as arcD gene, specifies a 53-kDa protein with arginine/ornithine exchange activity. Actually, this transporter was well examined in other bacteria like Pseudomonas aeruginosa and was suggested to play a crucial role in balancing the ornithine accumulation and the shortage of arginine in the cell during the arginolysis process by ensuring the cross-membrane exchange of these two substrates [66, 67]. According to MEMSAT-SVM analysis, this kind of transporter detected in all M. iowae strains was predicted to cross the membrane 13 times and thus to have a transmembrane localization. This bioinformatics result perfectly corroborates the topological model of the ArcD protein in the cytoplasmic membrane, which has been proposed more than 30 years ago based on a hydrophobicity plot and the inside-positive rule [67]. This similarity in protein sequence and topology allows us to theoretically hypothesize the same role for this transporter in this avian mycoplasma species consisting in the uptake and transportation of arginine and ornithine from/to the mycoplasmal cell. Nevertheless, complementary experimental biochemistry analysis would be necessary to confirm such an assumption. Interestingly, BLAST search showed that among all the avian mycoplasmas, only M. iowae species harbours this type of transporter. The existence of this arginine-ornithine exchange system in M. iowae might have resulted from a horizontal gene transfer (HGT) event since it has not been found in other species from the Pneumoniae and Hominis groups but in many Mollicutes belonging to the Spiroplasma group [63] or in other non-Mollicutes bacteria like Pseudomonas aeruginosa [66, 67].

The location of some genes encoding for enzymes involved in energetic metabolism on genomes of the three strains; M. meleagridis ATCC 29294, M. meleagridis MM_26B8_IPT, and M. gallinarum Mgn_IPT was graphically visualized in Fig. 5.

Fig. 5.

Circular visualization of the genome architecture of Mycoplasma meleagridis type strain ATCC 25294, Mycoplasma meleagridis field strain MM_26B8_IPT, and Mycoplasma gallinarum field strain Mgn_IPT. From inside to outside, the first circle shows the GC skew (inward projections in purple and outward projections in gold). The second circle shows the distribution of rRNA, tRNA, and tmRNA genes (in grey), while the third circle shows the CDSs (in light blue). Bacterial replication initiator (dnaA) (in black) and some genes related to virulence (in red and pink) and to energy metabolism (in green) were reported. MIB-MIP (Mycoplasma immunoglobulin binding-Mycoplasma immunoglobulin protease), licD (putative lipopolysaccharide diphosphonucleoside choline transferase), arcA (arginine deiminase), arcB (ornithine carbamoyltransferase), arcC (carbamate kinase), pgi (glucose-6-phosphate isomerase), pfk (6-phosphofructokinase), fba (fructose biphosphate aldolase), tpiA (triose-phosphate isomerase), gap (glyceraldehyde 3-phosphate dehydrogenase), pgk (3-phosphoglycerate kinase), pgm (phosphoglycerate mutase), eno (enolase), pyk (pyruvate kinase), pdhA (pyruvate dehydrogenase E1 α subunit), pdhB (pyruvate dehydrogenase E1 β subunit), pdhC (dihydrolipoamide acetyltransferase E2), pdhD (dihydrolipoamide dehydrogenase E3), rpiB (ribose-5-phosphate isomerase), rpe (Ribulose 5-phosphate 3-epimerase), tkt (transketolase). The fourth circle shows the position coordinates for each genome sequence. The three genomes are not complete, they are fragmented in contigs. This circular presentation, generated by Artemis DNAPlotter software, is a visualization of assembled contigs. Although this circular DNA map does not represent a complete genome version of one single chromosome, it can give a good estimation for functional annotation as well as an approximate position of genes

Deciphering of some virulence determinants in pathogenic avian mycoplasmas

Thanks to the numerous genome sequencing projects conducted during the past years in the mycoplasmology field, several genes were predicted to be associated with virulence, and experimentally verified in some cases, especially for the two avian mycoplasmas M. gallisepticum and M. synoviae, always considered as the most virulent and threatening species in avian intensive farming [68]. Pathogenicity of avian mycoplasmas was previously defined to be the combined actions of various components of their antigenically and functionally dynamic surface architecture. Such a coordinated action helps the pathogen in immune evasion, cytadherence, and colonization of its host [69]. The major clinical signs seen during avian mycoplasmosis consist mainly in respiratory disease, sinusitis, and synovitis. Chronic and asymptomatic infections are also common and they are more menacing due to their negative economic impact on the poultry production [70, 71]. The most studied virulence determinants in avian mycoplasmas are the cytadhesines and haemagglutinins; major surface proteins essential for adherence, colonization, and establishment of the disease in the host [26, 70]. In contrast, there are only few reports dealing with other components that could interact with the host extracellular matrix and contribute to the pathogenicity process [72]. Many virulence determinants whether for the well-known pathogens M. gallisepticum and M. synoviae or for the other avian mycoplasmas remain not fully elucidated. Little is known regarding M. meleagridis, M. anatis, M. iowae, and M. imitans virulence factors and their clinical presentations despite being important avian pathogens. For other avian mycoplasma species like M. gallinarum and M. gallinaceum, the pathogenic potential is not approved unanimously since they are still considered avirulent even though they were frequently isolated from diseased birds [73, 74]. For M. columbinum and M. columborale species, their isolation from both healthy and diseased pigeons suggests that they may be opportunistic pathogens at the most [75, 76].

Delving more deeply into the genomes of all these avian mycoplasma species would be very informative regarding the identification of their virulence genetic supports. To that end, we have conducted a comparative genomic analysis between the 40 avian mycoplasma strains enrolled in this study based on previous reports and studies. Results of this survey of genes associated to virulence are detailed below in different sections and a sum up of this research is presented in Table 2. In particular, some virulence associated-genes in our three genomes (M. meleagridis type strain 17529, M. meleagridis field strain MM_26B8_IPT, and M. gallinarum field strain Mgn_IPT) were spotlighted on the circular genome map presentation (Fig. 5).

Table 2.

Virulence factors-based comparison of forty avian mycoplasma strains

Avian mycoplasmas Species / Strain	Membrane-associated proteins					Nucleases		Proteases		MIB-MIP	NEAC		Metabolic pathways				Oxydative stress
	Cytadhesins					Total number	SNc	Total number	CysP
	gapA	crmA	crmB	crmC	licD						nanH	nanA	lpd	glp			osmC	ohr
														F	K	O
Ma / 1340	-	-	-	-	2	25	-	16	-	1/1	-	-	1	1	-	1	-	-
Ma / NCTC10156	-	-	-	-	2	25	-	16	-	2/1	-	-	1	1	-	1	-	-
Mco / NCTC10179	-	-	-	-	-	27	-	14	-	1/1	-	-	1	1	1	1	-	-
Mgc / B2096 8B	-	-	-	-	-	8	-	7	-	1/1	-	-	1	-	-	-	-	-
Mgc / NCTC10183	-	-	-	-	-	23	-	13	-	1/2	-	-	1	-	-	-	-	-
Ms / 53	-	-	-	-	-	18	-	14	1	4/2	1	1	1	-	-	-	-	-
Ms / WVU 1853T	-	-	-	-	-	23	-	9	1	4/2	1	1	1	-	-	-	-	-
Ms / MS-H (Chicken)	-	-	-	-	-	23	-	13	1	4/1	1	1	1	-	-	-	-	-
Ms / MS-H (Vaccine)	-	-	-	-	-	23	-	13	1	4/1	1	1	1	-	-	-	-	-
Mgn / Mgn_IPT	-	-	-	-	2	22	-	15	1	1/1	-	-	1	-	-	-	-	-
Mci / NCTC10178	-	-	-	-	-	25	-	14	-	1/1	-	-	1	-	-	-	-	-
Mci / SF7	-	-	-	-	-	17	-	10	-	1/1	-	-	1	-	-	-	-	-
Mm / ATCC 25294 (17529)	-	-	-	-	-	22	-	11	1	1/1	-	-	-	-	-	-	-	-
Mm / MM_26B8_IPT	-	-	-	-	-	21	-	12	1	1/1	-	-	-	-	-	-	-	-
Mm / NCTC10153	-	-	-	-	-	24	-	11	1	1/1	-	-	-	-	-	-	-	-
Mm / IZSVE/2944/9/2011	-	-	-	-	-	16	-	13	1	1/1	-	-	-	-	-	-	-	-
Mio / 695	-	-	-	-	-	20	-	15	-	-a	-	-	1	1	1	1	1	1
Mio / NCTC10185	-	-	-	-	-	22	-	18	-	-a	-	-	1	1	1	1	1	1
Mio / DK-CPA	-	-	-	-	-	14	-	14	-	-a	-	-	1	1	1	1	1	1
Mi / ATCC 51306	1	1	1	1	-	18	1	15	-	3/2	-	-	1	1	-	1	1	1
MG / Field strain Rlow	1	1	1	1	-	22	1	19	1	2/4b	1	-	1	1	1	1	1	1
MG / Field strain Rhigh	-	1	1	1	-	22	1	19	1	5/4b	1	-	1	1	-	1	1	1
MG / Field strain F	1	1	1	1	-	21	1	18	1	5/4b	1	-	1	1	1	1	1	1
MG / S6	1	1	1	1	-	24	1	20	1	4/4b	1	-	1	1	-	1	1	1
MG / VA94_7994-1-7P	1	1	1	1	-	19	1	19	1	4/4b	1	-	1	1	1	1	1	1
MG / NC95_13295-2-2P	1	1	1	1	-	19	1	19	1	4/4b	1	-	1	1	1	1	1	1
MG / NC96_1596-4-2P	1	1	1	1	-	19	1	18	1	4/4b	1	-	1	1	1	1	1	1
MG / NY01_2001.047–5-1P	1	1	1	1	-	18	1	18	1	4/4b	1	-	1	1	1	1	1	1
MG / WI01_2001.043–13-2P	1	1	1	1	-	19	1	19	1	4/4b	1	-	1	1	-	1	1	1
MG / NC06_2006.080–5-2P	1	1	1	1	-	16	1	18	1	4/4b	1	-	1	1	1	1	-	1
MG / CA06_2006.052–5-2P	1	1	1	1	-	19	1	19	1	3/4b	1	-	1	1	1	1	1	1
MG / NC08_2008.031–4-3P	1	1	1	1	-	16	1	18	1	4/4b	1	-	1	1	1	1	1	1
MG / K6356	1	1	1	1	-	19	1	16	1	4/3b	1	-	1	1	1	1	1	1
MG / K6372	1	1	1	1	-	19	1	13	1	4/3b	1	-	1	1	1	1	1	1
MG / K5322C	1	1	1	1	-	19	1	15	1	4/3b	1	-	1	1	1	1	1	1
MG / K6112B	1	1	1	1	-	19	1	16	1	4/3b	1	-	1	1	1	1	1	1
MG / K2966	1	1	1	1	-	19	1	16	1	4/3b	1	-	1	1	1	1	1	1
MG / K6222B	1	1	1	1	-	20	1	16	1	4/4b	1	-	1	1	1	1	1	1
MG / K6216D	1	1	1	1	-	20	1	16	1	4/4b	1	-	1	1	1	1	1	1
MG / K6208B	1	1	1	1	-	20	1	16	1	4/4b	1	-	1	1	1	1	1	1

Underlined strains are those sequenced by our team. The absence of virulence genes is signaled by the minus sign (-)

gapA cytadhesin protein GapA domain protein, crm cytadherence-related molecule, licD lipopolysaccharide diphosphonucleoside choline transferase, SNc staphylococcal nuclease (MGA_0676), CysP Cysteine protease (MGA_1153), MIB-MIP Mycoplasma Immunoglobulin Binding—Mycoplasma Immunoglobulin Protease, nanH neuraminidase, nanA N-acetylneuraminate lyase, lpd dihydrolipoamide dehydrogenase, glp glycerol permease, osmC osmotically inducible protein C (OsmC)-likeprotein, ohr organic hydroperoxide reductase

Ma Mycoplasma anatis, Mco Mycoplasma columborale, Mgc Mycoplasma gallinaceum, Ms Mycoplasma synoviae, Mgn Mycoplasma gallinarum, Mci Mycoplasma columbinum, Mm Mycoplasma meleagridis, Mio Mycoplasma iowae, Mi Mycoplasma imitans, MG Mycoplasma gallisepticum

^a In M. iowae, the MIB-MIP system was not found but the MIB-related protein M was identified

^b In M. gallisepticum, several copies of the MIB-MIP system were predicted as well as one copy of the gene encoding protein M

Membrane-associated proteins (cytadhesins, haemagglutinins, and lipoproteins)

The mycoplasma membrane components are principally proteins (over two-third) and lipids (one-third) [77]. In contrast to their limited number in membranes of other bacteria, membrane lipoproteins are remarkably abundant in mycoplasmas [78]. Membrane lipoproteins belong to the most dominant antigens in mycoplasmas and they are known to undergo antigenic variation (phase and/or size variation), which enables evasion of the host immune system [15]. All of pathogenic mycoplasmas firmly adhere to epithelial surfaces of their host tissues and are usually non-invasive organisms. However, some reports have evoked the ability of these microorganisms to invade different cell types [79]. With regards to avian mycoplasmas, cell invasion was also reported in some species like M. gallisepticum by showing through experiments its ability to invade nonphagocytic cells such as chicken embryonic fibroblasts and chicken erythrocytes [80, 81]. The adherence of mycoplasmas to their host cells is a prerequisite for successful colonization and thus for the initiation of the pathogenesis processes resulting in cell alterations. Effectively, in some mycoplasmas like M. pneumoniae, it was proven that the loss of adhesion capacity results in a loss of infectivity and vice versa (infectivity and virulence were recovered with reversion of cytadhering phenotype) [82]. The cytadherence process in mycoplasmas is multifactorial. The attachment is mediated by specific interactions between mycoplasma cytadhesins and their corresponding host-cell receptors [15, 83, 84]. This phenomenon was brought out in the avian mycoplasma M. gallisepticum through the identification of a flask-shaped organelle. This structural feature, also named tip, was previously well described in the human pathogenic mycoplasmas M. pneumoniae and M. genitalium [28]. The attachment process of M. gallisepticum virulent strain Rlow to the epithelial tissues of its host depends mainly on a coordinate action between a major cytadhesin GapA and several accessory cytadhesin-related molecules such as CrmA, CrmB, and CrmC [8, 85]. According to our analysis, the genes (MGA_RS01025, MGA_RS01030, MGA_RS01035, and MGA_RS01040) encoding for the aforementioned proteins were found in M. gallisepticum strain Rlow genome. These genes were also detected in the other M. gallisepticum strains studied here. In particular, in the GeneBank report of the strain Rhigh (attenuated derivative of strain Rlow), it was noticed that the gapA gene is disrupted. This finding is in agreement with the frameshift mutation detected in this gene explaining its disruption and thus the lack of GapA expression [86]. The absence of a such primary cytadhesin protein, in addition to loss and/or shift away of other proteins, is responsible of the attenuated Rhigh phenotype [10]. Regarding the other avian mycoplasmas, the set of genes gapA, crmA, crmB, and crmC was absent in all the species excepting M. imitans (P690_RS0103325, P690_RS0103320, P690_RS0103315, and P690_RS0103310, respectively). This was not astounding since M. imitans is a pathogenic mycoplasma species and very close phylogenetically to M. gallisepticum.

Furthermore, it has been well described before that M. gallisepticum and M. synoviae have a large gene family encoding lipoproteins that function as haemagglutinins [87–90], referred to as pMGA [87, 88] and vlhA [89, 90], respectively. Despite the high number of pMGA genes in each M. gallisepticum strain, only one gene is transcribted. The remaining copies are either transcriptionally silent or transcribed at insignificant levels. Thus, only one pMGA molecule is expressed at once per strain [87], which explains the largeness of pMGA multigene families (32—70 genes) in M. gallisepticum species [88]. A similar phenomenon was also observed in M. synoviae. In fact, the vlhA gene family is comprised of a unique vlhA expressed gene and many pseudogenes, all confined to the same region in the M. synoviae genome. The expressed version of this gene encodes for a haemagglutinin involved in the pathogenicity. This haemagglutinin is endowed with antigenic variation resulting from recombination events through which the vlhA pseudogenes are recruited and expressed, thus generating variants of the initial vlhA expressed version [90]. This gene multitude was proved by our in silico screening of M. gallisepticum and M. synoviae genomes showing numerous variant haemagglutinins encoding genes in the strains of these two species (data not shown). These immunodominant variably expressed lipoproteins were proven to be capable of undergoing phase variation and are thought to be involved in establishment of chronic infection through immune system evasion [87–90]. The existence of immunogenic and variably expressed surface proteins was hypothesized before for other avian mycoplasma species such as M. meleagridis, M. iowae, and M. imitans, but their role as haemagglutinins was doubtful and contrevertial [26]. A few old studies have reported the ability of some of these avian mycoplasma species (two isolates of M. meleagridis named E2 and 8M92 and one serotype J of M. iowae) to induce hemagglutination of red blood cells [91, 92]. In literature, no particular protein or gene has been linked to this activity in M. meleagridis. As for M. iowae, the output of a Southern blot technique showed that the DNA of serotype J of this species bounded to a specific pMGA1.2 gene probe [26, 92], which proposes the ability of this avian mycoplasma species to aggluntinate erythrocytes. As for M. imitans, a gene family showing homology to the pMGA and vlhA gene families was identified in the type strain ATCC 51306 [26, 92]. This finding speculated the horizontal transfer of this gene family between mycoplasmas sharing avian hosts since all of M. gallisepticum, M. synoviae, and M. imitans are infecting gallinaceous species [26, 92]. Based on the above data, we scrutinized M. meleagridis, M. iowae, and M. imitans genomes studied herein for haemagglutination-related sequences. BLASTP analysis revealed no matches. For M. meleagridis and M. iowae, this could be explained by the fact that we are using different strains from those in which hemagglutination activity was suggested. For M. imitans, the non detection of any homolog for a hemagglutination gene was unexpected since we are working with the same strain in which this gene has been previously detected. This may be due to the low quality of the genome sequence of this strain (it is a scaffold-level genome) which could affect the annotation accuracy (indeed, 149 among 772 genes were annotated as hypothetical proteins). Furthermore, to make the analysis more complete, we have extended our screening for haemagglutinin genes in all the other avian mycoplasma strains. Similarly, we did not find any relevant result.

Moreover, many other membrane-associated proteins were classified as virulence factors as they were strongly implicated in host–pathogen interaction. One of these are the outer membrane protein LicD (putative lipopolysaccharide diphosphonucleoside choline transferase). This protein contributes in the control of incorporation of environmental choline into lipopolysaccharide, essential component of the outer membrane of Gram-negative bacteria, which results in the formation of phosphorylcholine (ChoP). In some bacteria like Haemophilus influenza and Haemophilus haemolyticus, it has been shown that polymorphisms in the central portion of the LicD-encoded proteins contribute in the modification of phase-variable ChoP [93]. From these data, licD gene has been suggested to aid in improvement of adherence ability of bacteria to epithelial cells and, hence, to contribute in pathogenicity. CDSs homologous to LicD were detected before in mycoplasmas, but only in few murine and human mycoplasma species (M. pulmonis, M. arthritidis, and M. fermentans) [94–97]. The function of licD gene, also named mf1 in M. fermentans, was experimentally tested and has shown the importance of the phosphocholine transfer step in the glycoglycerophospholipid biosynthesis pathway of M. fermentans [96]. Another study on this same mycoplasma species has suggested that the variation in mf1 gene sequence can contribute to the pathogenic process [97]. In avian mycoplasmas, two copies of the licD gene (GIG_00075 and GIG_01785) were previously found in the M. anatis type strain 1340 and were identified as putative virulence determinants [11]. Through our comparative genomic analysis, we have confirmed this finding and we have also revealed the presence of this couple of genes (DP067_RS00970 and DP067_RS02650) in the other strain of M. anatis (NCTC10156) enrolled in our study. Homolgous of these genes (MGALLINA_RS00220 and MGALLINA_RS01285), both encoding for diacylglycerol cholinephosphotransferase Mf1, were also detected in M. gallinarum field strain Mgn_IPT. BLASTP analysis of these proteins revealed that they harbour the LicD family conserved domain (pfam04991) associated with ChoP metabolism. Either in M. anatis or M. gallinarum, the two licD, or mf1, alleles are distant in the genome and their surrounding sequences are different. This probably indicates that these copies have been randomly duplicated. As for the other avian mycoplasma strains, this couple of genes was totally absent. There is no clear scenario that can explain the origin and the exclusive existence of this couple of genes only in the two avian mycoplasma species M. anatis and M. gallinarum, especially that they are not sharing the same host and that M. gallinarum is known to be apathogenic. To our knowledge, no previous experimental report underlying the possible role of LicD protein in virulence has been previously published with regards to any avian mycoplasma.

Nucleases and proteases

All avian mycoplasma strains, included in this comparative analysis, were found to harbour many nucleases and proteases (Table 2). The omnipresence of these enzymes in avian mycoplasma species is consistent with the limited biosynthetic capacity often reported within the Mollicutes class bacteria [15]. Indeed, because of their reduced genome, the mycoplasmas are inherently deficiencient in terms of metabolic pathways leading to a complete dependence on their hosts to acquire essential nutrients for their survival such as nucleotides, some aa, fatty acids, and sterols [15]. Degradation of the host nucleotides via nucleases was reported to be a key source of nucleic precursors for mycoplasmas [98]. Likewise, there are no complete routes for aa synthesis in mycoplasmas and the possible way to acquire these monomers is to bring them in either from their hosts or from culture media [9]. Importation of exogenous peptides into the cell is mediated through membrane transporters. The bacterial oligopeptide permease (Opp) transport system is an example and the ATP-binding cassette (ABC) transporters are the most prominent in mycoplasmas [99–101].

Apart the importance of nucleases and proteases in the life cycle of mycoplasmas and in their vital functions such as DNA repair, replication, transcription, translation, and protein maturation, it has been reported that oligonucleotide uptake and protein degradation could also be associated with host adaptation and pathogenicity in these microorganisms [99, 102–105].

In avian mycoplasmas, involvement of a nuclease in pathogenicity was reported experimentally only once. This nuclease was first identified in M. gallisepticum strain Rlow at the locus tag MGA_0676 (currently tagged as MGA_RS00165 and annotated as a thermonuclease family protein). Its characterization has shown that it is a 276 aa-protein associated to the membrane, harbouring a staphylococcal nuclease SNc conserved domain (accession number: cd00175), and having an enzymatic activity dependent on Ca²⁺ ions. A cytotoxic effect was attributed to this nuclease as it was found to be able to induce apoptosis and to cause pathological effects in chicken cells [106]. BLASTP of the aa sequence of this nuclease revealed its existence (with full coverage and identity percentages ranging from 98.8 to 100%) in several other M. gallisepticum strains among those studied here, whether they were isolated from poultry birds or house-finches. A similar copy (100% coverage, 74% identity) of this enzyme was also detected in the close phylogenetic neighbor species M. imitans (P690_RS0103095: thermonuclease family protein). As for the remaining avian mycoplasma species, this nuclease was absent. Nevertheless, the lack of detection of this nuclease does not exclude the possible existence of other nucleases with potential incrimination in pathogenicity that are not discovered and studied yet. In M. meleagridis type strain for example, a nuclease named Mm19 (MMELEA_RS00585) was molecularly and functionally well characterized. It was shown to be specific to this species among the avian mycoplasmas group as it was not detected in any of the other strains studied here, but no data regarding its pathogenic effect was reported [29].

Like other microbial pathogens, mycoplasmas utilize proteases as virulence factors, which contributes to invasion and survival in their hosts and may largely impact their pathogenicity [105, 107, 108]. One proof of this is the ability of protases to degrade immunoglobulins (IgG) and components of the complement system of hosts during the infection propagation as a means to evade the immune system. Effectively, it was demonstrated that a cysteine protease (CysP) expressed in M. gallisepticum strain Rlow (MGA_1153, currently tagged MGA_RS01740) and M. synoviae 53 strain (MS53_0590, currently tagged MS53_RS03095) is capable of cleaving chicken IgG into antigen-binding fragment (Fab) and crystallizable region fragment (Fc) [109]. We have detected this protease in all M. synoviae strains studied here (MS53_RS03095, VY93_RS03235, MSH_RS03060, and mshv_RS01700). As for M. gallisepticum, strains Rlow and S6 were the only strains with complete genome in which this peptidase was carried in one gene. In the other strains, a couple of juxtaposed genes named pepC_1 and pepC_2 seems to encode for homologous of this peptidase (Example in the strain Rhigh: MGAH_RS02525 and MGAH_RS02530). Moreover, homologs of this peptidase were detected in M. gallinarum strain Mgn-IPT (MGALLINA_RS01865) and in the four strains of M. meleagridis (MMELEA_RS00650, MM26B8_RS00635, EXC33_RS02155, and ASB56_RS01110). All these proteases were found to harbour a peptidase_C1 domain (accession cd02619). These in silico findings may reflect the ability of these two species (M. gallinarum and M. meleagridis) to cleave the IgG of their hosts too. However, in vitro/in vivo studies are needed to prove this. In case of confirming the functionality of this enzyme, particularly in M. gallinarum, this may considerably help predicting the pathogenic potential of mycoplasmas. Regarding the other avian mycoplasma species, this protease was not found in any of them. Generally, proteases can cause tissue destruction and inactivation of host defense molecules. It was also reported that in some bacterial pathogens, some toxins need proteases to be activated [99]. In avian mycoplasmas, no study has yet treated any of these latter pathogenicity aspects and no investigation has unlocked their mechanisms.

Immunoglobulin blocking proteins

Mycoplasmas have specifically evolved a family of immunoglobulin-blocking proteins, which have the ability to bind to the antibodies light chain and to subsequently disrupt their antigen-binding function [110]. Ig-blocking proteins are extracellular, surface anchored, and split in two sub-groups of effectors. The first is comprised of MIB (Mycoplasma Immunoglobulin Binding) homologs, a set of proteins that have the ability to bind to either free antibodies or antibodies forming immune complexes with their antigens. Upon binding, MIBs disrupt the conformation of the variable domains of Ig heavy and light chains, V_H and V_L, preventing free antibodies from binding their antigen or forcing the separation of immune complexes. Interestingly, MIB has the ability to recruit MIP (Mycoplasma Immunoglobulin Protease), a serine protease that has the ability to cleave off the V_H domain, resulting in a broken antibody that can no longer recognize its target [110, 111].

The prototypical MIB&MIP have been characterized in the caprine pathogen M. mycoides subsp. capri, and are conserved in the majority of the animal pathogenic mycoplasma species. Avian mycoplasmas are no exception, and we have identified putative MIBs and MIPs in 37 of the 40 strains studied here (Table 2). The number of MIBs and MIPs per genome vary greatly, ranging from one of each in M. columborale, M. gallinaceum, M. gallinarum, and M. columbinum to four or five copies in most M. gallisepticum strains. It is notable that the loci encoding these multiple MIBs and MIPs are often scattered across the genomes, in stark contrast to what is observed in members of the Mycoides cluster where MIBs and MIPs coding sequences are mostly found in operons. It should be noted that the number of MIBs and MIPs per genome does not appear to be correlated with virulence, and that the presence of multiple variants is probably linked to an antigenic variation mechanism [112].

The second sub-group of Ig-blocking proteins is comprised of homologs of “Protein M”, a set of MIB derivatives that have maintained their ability to bind to the antibody light chain through a conserved domain, but have lost the ability to interact with and dissociate immune complexes. Instead, “Protein M’s” appear to extend their C-terminal domain over the antigen-binding to sterically block it and prevent the formation of immune complexes. The in vivo function of “Protein M” remains cryptic, and it also appears to have lost its ability to partner with an Ig-protease as no homolog of MIP could be found to be genetically associated with it. This proteolytic role could nonetheless be attributed to one of the many surface-anchored proteases found in mycoplasmas [110–112]. “Protein M’s” were initially characterized in the closely related species M. genitalium and M. pneumoniae, and appear to be a specificity of the Pneumoniae group. Distant homologs are predicted in the three M. iowae strains studied here, as well as in all M. gallisepticum strains. This co-occurrence of MIB and Protein M is atypical, as in all other known mycoplasmas, these two systems are mutually exclusive. It is also notable that the emergence of “Protein M” in replacement of MIB-MIP seems to have occurred independently in two distinct phylogenetic branches, and in species with drastically divergent hosts (human and avian) [110–112].

At this point, no study has been reported on these systems in avian mycoplasmas. In addition, the exact function of Ig-blocking proteins during colonization and infection remains nebulous, although these systems clearly appear to be involved in evasion of the humoral immune response.

Neuraminidase enzymatic activity (NEAC)

Neuraminidases, also called sialidases, are glycosyl hydrolase enzymes present in bacteria and viruses [113]. In bacteria, including mycoplasmas, NEAC is used to recognize host cells, colonize and disseminate within the host, degrade the extracellular matrix and scavenge sialic acids as nutrients [114–117]. Thus, NEAC was thought to constitute a potential enzymatic basis for virulence in many pathogenic microorganisms. In Streptococcus pneumoniae for example, some sialidase-associated genes were proved critical during sepsis and infection of the respiratory tract (nasopharyngeal tract as well as upper and lower respiratory tract) [118]. In mycoplasmas, NEAC is considered to be relatively rare. This activity was reported since long time ago for the first time in the avian pathogen M. gallisepticum [119, 120]. Then, some subsequent reports have shown its existence in other species isolated from various hosts such as M. alligatoris [117], M. pneumoniae [121], M. bovis [122], M. synoviae [123], and M. neurolyticum [124]. More recently, WGS as well as functional studies have revealed some of the genetic basis of NEAC in some mycoplasma strains. With regards to avian mycoplasmas for instance, M. gallisepticum strain Rlow and M. synoviae strain 53 were shown to share a gene named sialidase nanH encoding for neuraminidase enzyme [8, 9]. Through the present in silico analysis, we have first confirmed the presence of this gene (for which the product was named sialidase family protein) in M. gallisepticum strain Rlow (MGA_RS03180, old locus tag MGA_0329) and in M. synoviae strain 53 (MS53_RS01080, old locus tag MS53_0199), and then we have checked its presence in the other strains of these two species. Many of them were found to harbour a copy of this nanH gene with full coverage and very high similarity percentages exceeding 95%. Moreover, a gene named nanA encoding for acylneuraminate lyase and located directly upstream of the nanH gene was also spotted, but only in M. synoviae strains (MS53_RS01075, VY93_RS01140, MSH_RS01140, and mshv_RS03400). Likewise, a series of other tandemly arranged genes associated with sialic acid scavenging and degradation pathway were detected in the upstream region of nanH and nanA genes in M. synoviae strains. This organization could be in favor of the expression and regulation of this set of genes under the control of the same promoter. This result is in line with what has been reported before, in that in many bacterial genomes the silidase gene is usualy located close to other genes encoding enzymes involved in the sialic acid pathway either through facilitating its transport or its metabolism [115]. It is worth mentioning that only M. synoviae (represented by its 4 strains) was found to carry this genes’ cluster among the tested avian mycoplasmas. Thus, one could suggest that NEAC in this avian mycoplasma is probably more important than in the other avian mycoplasmas. For M. synoviae field strain 53, no experimental proof of sialidase activity nor clinical data have been reported so far. However, this was not the case for the type strain of M. synoviae WVU 1853^T for which considerable NEAC level was detected [123]. None of the remaining avian mycoplasma strains was found to harbour putative coding sequences involved in sialic acid catabolism, which suggests the presence of NEAC in them. Opposingly, it was shown in a previous experimental study that M. meleagridis type strain ATCC 25294 and M. iowae type strain 695 exhibit NEAC [125]. As for M. gallinarum and M. anatis, this same study indicated that these species are deprived of NEAC. This is in concordance with the lack, in their genomes, of nanH and nanA genes and all the other genes encoding for enzymes involved in sialic acid metabolism pathway.

Moreover, association of NEAC with virulence was proved in M. gallisepticum and M. synoviae. For the latter species, it was found that strains isolated from clinically symptomatic birds exhibit more NEAC than strains isolated from asymptomatic birds. For M. gallisepticum strain Rlow, chickens infected with the knockout mutants lacking sialidase activity, created by disrupting the nanH gene, had significantly less severe tracheal lesions than those infected with strain Rlow wild type [123, 126]. Generally, NEAC exists in some avian mycoplasmas but with variable levels of expression between species and even between strains of a same species. Correlation between genetic characteristics and phenotypic analysis is important to properly evaluate the presence of NEAC in avian mycoplasmas as well as its effective involvement in pathogenicity.

Virulence-associated metabolic pathway genes

Proteins implicated in metabolic pathways are traditionally not considered to be associated with virulence. Yet, some studies have contradicted this assumption and have investigated the role of these proteins in bacterial pathogenicity. It is known that carbon and energy are crucial for bacteria to grow and replicate. These are usually derived from the environment through the catabolism of more-complex organic molecules such as carbohydrates, lipids, and proteins. For pathogenic bacteria, carbon and energy are acquired from the host organism either parasitically or destructively. This is achieved in part by the synthesis of virulence factors that are able to kill host cells and thus to catabolize macromolecules. Hence, it seems not surprizing to find that the regulation of many virulence determinants could be dependent on nutrient availability [127]. This assumption could be valid for mycoplasmas since they are known to be limited in their metabolic arsenal, which makes the close association with the host cell crucial for acquisition of nutrients [128]. In avian mycoplasmas, some genes coding for proteins involved in metabolism have been shown to contribute to virulence. This is typically the case of the lpd gene encoding for the dihydrolipoamide dehydrogenase (Lpd), which is a component of the pyruvate dehydrogenase complex (PDC), a central carbohydrate metabolism in fermentative mycoplasmas (described above in energy-yielding metabolisms section). In M. gallisepticum for example, it was proven that Lpd is required for in vivo growth and survival of this species in its host [10, 129]. The contribution of lpd gene (MGA_RS02750) in pathogenicity of M. gallisepticum strain Rlow was tested and confirmed by observing the effect of its altered-sequence expression on the host cells. It was shown that the ability of this virulent strain to cause tracheal lesions was reduced because of a mutation in the lpd gene. Disruption of Lpd protein expression in M. gallisepticum resulted in no acetyl-CoA production and, therefore, no ATP is released at the end of glycolysis. This leads to an intracellular energy shortage in M. gallisepticum affecting its capacity to uptake and transport precursor molecules necessary for its survival. It was confirmed that this deficiency is able to affect the M. gallisepticum virulence potential and its ability to persist and adapt to its host environment [129]. Alike Rlow strain, a single copy of lpd gene was detected in many other M. gallisepticum strains and also in the strains of the other glycolytic avian mycoplasma species. Remarkably, this gene was identified in M. gallinarum (MGALLINA_RS03325) and in both M. columbinum strains (EXC37_RS02430 and MCSF7_RS02520) despite being nonglycolytic. Aside from M. gallisepticum Rlow strain, no experiment has been done before to test the lpd gene involvement in virulence.

Another example linking metabolic activities to virulence is the cytotoxic effect caused by hydrogen peroxide (H₂O₂) resulting from glycerol metabolism in mycoplasmas [130]. In fact, many fermentative mycoplasmas are able to utilize glycerol as alternative carbon source to glucose. In these mycoplasmas, a set of three glp genes encoding for proteins involved in glycerol transport and metabolism was defined: glpF gene is implicated in transportation of glycerol into the cell, glpK in phosphorylation of glycerol to glycerol-3-phosphate, and glpO in conversion of glycerol-3-phosphate to dihydroxyacetone phosphate and H₂O₂ [131]. For some mycoplasma species such as M. gallisepticum, it was shown that the wild-type strain Rlow is capable of producing H₂O₂ when grown in glycerol and is cytotoxic to cells in culture. In contrast, this cytotoxic effect was aborted in glp mutants, demonstrating that cytotoxicity is tightly linked to the catabolism of glycerol [132]. The three glpFKO genes were detected in M. gallisepticum strain Rlow [8] and in almost all the other M. gallisepticum strains. They were also found in M. iowae and M. columborale among the rest of avian mycoplasmas. This probably reflects the ability of these species to catabolize glycerol, to produce H₂O₂, and possibly to be toxic to their hosts. In contrast, no trace of proteins involved in the uptake and conversion of glycerol was detected in the remaining avian mycoplasma species.

Through the examples listed above, the role of metabolic enzymes as primary virulence factors is suggested. This could also be a form of adaptation since the virulence of pathogenic mycoplasmas is not mainly determined by toxins, invasins, and cytolysins like the other pathogenic bacteria [133]. In fact, it is known that the adaptation to an altered environment usually activates the expression of virulence genes, which explains that several virulence genes are under the control of environmental nutrients [127]. This may reflect the evolution of microbial pathogenicity, in general, including mycoplasmal pathogens [133].

Oxidative stress

Depending on the host adaptation and immune response, the environment of the site of mycoplasmas infection is perpetually changing. This explains how these microorganisms are continuously exposed to different types of assaults and stresses such as heat shocks, osmolarity variations, hormone exposure, and oxidative stress enforced by hosts to protect themselves [134]. Here, we focus on the search of genes responsible for the resistance to oxidative stress in avian mycoplasmas. Particularly, the host-generated peroxides and reactive oxygen species (ROS), such as H₂O₂ and organic peroxides, are poisonous compounds capable of causing serious damages in various cellular macromolecules including DNA, proteins, and lipids leading to degrading bacteria [135]. It is important to note that apart from exogenous ROS (produced by the host as an immune response to infection), mycoplasmas should also resist to their own-produced ROS (endogenous) [136]. In several mollicutes, a significant role in virulence has been attributed to genes directly involved in carbon metabolism resulting in the formation of hydrogen peroxide [131]. This cytotoxicity factor has been well documented in the human pathogen M. pneumoniae [137, 138]. By detoxifying these noxious substances, mycoplasmas could ensure their survival in the host milieu and could then continue to exercise their pathogenic effect. One of the factors potentially involved in defense against oxidative stress is the osmotically inducible protein C (OsmC) [139]. Among avian mycoplasmas, this protein was previously found in M. gallisepticum. Functional characterization has shown that this protein can degrade both organic and inorganic peroxides, thus conferring a peroxide resistance in the extracellular environment and aiding in the attachment of M. gallisepticum to eukaryotic cells [140]. Furthermore, implication of OsmC in M. gallisepticum pathogenicity was previously evoqued since it was found to be a heparin-binding protein, and, hence might be involved in adhesion to host cells [72]. The osmC gene coding for the abovenamed protein was detected in strains of M. gallisepticum, M. iowae, and M. imitans. In addition, another gene named ohr (organic hydroperoxide reductase) encoding for an OsmC homolog was also found, indicating that peroxidase function in these avian mycoplasma species is important. Notably, this finding could be a reflection of a phylogenetic relatedness since all these species belong to the Pneumoniae group. No avian mycoplasma species included in the Hominis group was found to harbour these genes.

Conclusion

Here we scrutinized genomic data of forty strains belonging to ten different avian mycoplasma species, performing a myriad of intraspecies and interspecies comparative analyses based on five key aspects (genomic hallmarks, phylogeny, pan/core genome, energetic metabolism, and virulence). Phylogenetic analyses, based on concatenation of orthologs, confirmed the affiliation of avian mycoplasmas to two distinct groups in the Mollicutes class, with an obvious segregation into clusters. Yet, no linkage or co-evolution seem to exist between groups/clusters and host specificity. Avian mycoplasmas were found to present considerable intra- and inter-species variability and to have an open pan genome. However, the core genome does not appear to be affected by this variability. A restricted number of 150 essential gene clusters, coding for vital functions, was found to be shared between all avian mycoplasmas. Collectively, our data suggest that mycoplasma rely on more than one metabolism pathway to produce energy.

Regarding pathogenicity, avian mycoplasmas were found to be armed with different arsenals. Some of these, including diverse nucleases, proteases, and MIBs&MIPs were omnipresent in the avian mycoplasma species enrolled in this study. The existence of the Ig-blocking proteins in avian mycoplasma species was not an exception among mycoplasmas. Nevertheless, some singular features seem to render these systems unique in this group of mycoplasmas. First, the dispersion of genes encoding these proteins across the genomes contrasts with the organization in operon reported in the other mycoplasma species. Second, the co-occurrence of MIB and Protein M in some avian mycoplasma species such as M. iowae and M. gallisepticum, is atypical since these two systems are mutually exclusive in the other mycoplasmas. Some major virulence factors like the capacity to adhere to host cells were found to be exclusive to M. gallisepticum, M. synoviae, and M. imitans species, a fact that further points out their pathogenicity potential often described in literature. NEAC, whose exitence in mycoplasmas has rarely been reported, was effectively detected in only a few avian mycoplasma species. M. synoviae was remarkable in this regard given the high number of genes involved in the transport and metabolism of sialic acid found in this species.

Overall, our findings highlight the variability that characterizes the genomic make up of avian mycoplasmas, pointing to the diversity of their putative virulence factors, which underscores the hidden pathogenic potential of these minimalist microorganisms. The data garnered from these comparative genomic analyses would help designing rational experimental interventions in order to better decipher the molecular basis underlying the biology of avian mycoplasmas.

Materials and methods

Mycoplasma strains

Forty strains representing ten avian mycoplasma species (M. anatis, M. columborale, M. gallinaceum, M. synoviae, M. gallinarum, M. columbinum, M. meleagridis, M. iowae, M. imitans, and M. gallisepticum), were used in this study. Among this selection, the genome sequences of 37 strains are available in the GenBank and 3 genomes (M. meleagridis type strain 17529, M. meleagridis field strain MM_26B8_IPT, and M. gallinarum field strain Mgn_IPT) were sequenced by our team [12, 13]. Unlike M. meleagridis type strain 17529, the two other strains were both isolated in our laboratory from chickens with respiratory symptoms and poor performance, in the northeast of Tunisia. The assembly numbers and the main characteristics of all these avian mycoplasmas are presented in Table 1.

In silico comparative genomic analyses

Comparative genomic analyses of the avian mycoplasma genomes cited above were mainly performed using a combination of tools available in the MolliGen public database [141]. Genes and proteins were detected based on various in silico researches. BLAST program (megablast, blastp, blastn) was used to accomplish nucleotide and protein sequence alignments for homology searches in the NCBI database (https://www.ncbi.nlm.nih.gov/) with an expect threshold fixed at 0.05. Some other analyses related to the prediction of membrane helix and secondary structure were conducted using PSIPRED and MEMSAT-SVM methods available in the PSIPRED protein analysis workbench. Clustering of the putative proteins was achieved based on sequence homology [142]. A minimum of 50% sequence conservation over 50% of the protein/gene length was fixed as a threshold to define a conserved gene and to include it in an orthologous protein family or cluster [143]. The pan and core genomes were generated by gene number estimation curves [143, 144].

Phylogenetic tree construction

Phylogenetic analyses of the Mollicutes class were conducted using maximum likelihood method. A total of 31 Mollicutes species, including the 10 avian mycoplasma species were compared and the bacterium Bacillus subtilis was used as an outgroup species. The tree was built based on the concatenation of multiple sequence alignments of 47 single copy core genes identified using get_homologues (v.05032019) [145] with the COGtriangle [142] clustering algorithm. Each group of core protein sequences were aligned using Clustal Omega (v.1.2.1) [146]. Unaligned and low-confidence regions were removed from the alignment using Gblocks (v.0.91b) [147] and the alignments were then concatenated to produce a sequence matrix of 9449 amino acid sites. The CpREV model of evolution [148] was used for the tree construction as was determined using ProtTest (v.3.4.2) [149]. The tree was then created using RaxML (v.8.2.12) [150] using the GAMMA model of rate of heterogeneity, 150 bootstrap replicates were made using the autoFC bootstopping criterion to determine the number of replicates to perform. The names of the used genes and their encoded products are available in Molligen database (http://cbi.labri.fr/outils/molligen/), and they are listed in Additional file 1.

Circular presentation of mycoplasma genomes

Circular genome analysis was performed only for our three sequenced strains (M. meleagridis ATCC 25294, M. meleagridis MM_26B8_IPT, and M. gallinarum Mgn_IPT) to display some specific features related to virulence and metabolism. All three genomes are at contigs level, they are not circularized yet. Thus, contigs were first randomly assembled then annotated using Prokka (rapid prokaryotic genome annotation pipeline) [151]. GFF output files of Prokka were used as input for Artemis software [152]. DNAPlotter package of Artemis was employed to generate circular DNA map and to perform GC skew for each of the three selected strains. DNAPlotter is accurate for genome visualization of fragmented complex microbial genomes, which makes it suitable for an approximate gene positioning [153]. Genes of interest were tagged on the circular map to identify their position. Track manager was used to add CDSs and RNA genes to the maps. Position on the map of some genes that were not annotated using Prokka, was curated manually. Nucleotide sequences of these genes were downloaded in FASTA format and served to create a new database using makeblast [154]. Then, they were aligned to the assembled FASTA of our genomes to identify their exact position. The circular DNA diagrams inferred from this anlysis are just an approximate illustration of some genomic features in our three selected mycoplasma strains, they do not necessarily represent the reality of the full genomic context.

Abbreviations

CDS

Coding DNA sequence

EMP

Embden-meyerhof-parnas

PDH

Pyruvate dehydrogenase

PPP

Pentose phosphate pathway

ADI

Arginine dihydrolase

MIB

Mycoplasma immunoglobulin binding

MIP

Mycoplasma immunoglobulin protease

NEAC

Neuraminidase enzymatic activity

Authors’ contributions

Writing-original draft, E.Y. Methodology and figure preparation, E.Y., V.B., P.S.P., and S.C. Writing-review and editing, E.Y., V.B., P.S.P., Y.A., H.M., A.B., and B.B.A.M. Project administration, B.B.A.M. Funding acquisition, B.B.A.M. All authors have read and agreed to the final version of the manuscript.

Funding

This study was funded by the Tunisian Ministry of Higher Education and Scientific Research (MMVDB Project).

Data availability

No datasets were generated or analysed during the current study.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.