Introduction

Propionibacterium acnes is a commensal bacterium on human skin and mucosal surfaces and is considered causative of acne. Previous studies reported the isolation of P. acnes from several tissues, including the conjunctiva, external ear canal, oral cavity, upper respiratory tract and intestine1 and the possible association of P. acnes with inflammatory disease, such as chronic prostatitis2, endocarditis3 and sarcoidosis4,5,6.

Sarcoidosis is a systemic granulomatous disease with unknown etiology that seems to result from the exposure of a genetically susceptible subject to an environmental agent and microbial etiologies of sarcoidosis have long been considered based on the clinical similarity to infectious granulomatous diseases7. P. acnes is the only microorganism isolated from sarcoid lesions by bacterial culture to date8,9 and one of the most commonly implicated etiologic agents of sarcoidosis10,11. A series of Japanese studies proposed an etiology of sarcoidosis as an allergic endogenous infection caused by this indigenous bacterium.

According to the currently-proposed etiology of sarcoidosis10,11, this low-virulence bacterium causes latent infection in the lungs and lymph nodes and persists in a cell-wall-deficient form. This dormant form of P. acnes can be activated endogenously under certain environmental conditions and proliferate in cells at the site of the latent infection. In patients who are hypersensitive to this endogenous bacterium, granulomatous inflammation is triggered by intracellular proliferation of the bacterium. If a certain strain of P. acnes causes sarcoidosis, such a causative P. acnes strain may have some specific characteristics that confer its intracellular persistency, cell-wall-deficiency and endogenous activation, or there may be specific antigenicity of the bacterium in sarcoidosis patients.

Many studies of acne vulgaris report that P. acnes exhibits phenotypic and genotypic diversity12,13,14. In connection with sarcoidosis, Ishige et al. compared genotypes of P. acnes strains isolated from the lungs and lymph nodes with those of P. acnes indigenous to the skin, conjunctivae and intestine using random amplified polymorphic DNA analysis15. They found that P. acnes strains from a particular site were genetically similar, more so than isolates obtained from different sites. Moreover, Minegishi et al. recently determined the complete genome sequence of a P. acnes isolate (C1) from granulomatous inflammatory lesions of a patient with cutaneous sarcoidosis16.

In the present study, we initially performed core genome analysis and multiple genome alignment using the whole genome sequence from the C1 strain of P. acnes, compared with 76 and 9 strains of P. acnes from a public database, respectively, to search genetic profiles of P. acnes from sarcoid tissue samples. In addition, we examined 24 and 36 isolates from sarcoid and non-sarcoid tissue samples, respectively, by multilocus sequence typing (MLST) and polymerase chain reaction (PCR) detection for a P. acnes-specific insertion sequence (IS) and extrinsic protein-coding DNA sequences (CDSs) of a novel transposon. The roles of the P. acnes-specific transposon with novel ISs and the cell-invasiveness of P. acnes with the transposon are discussed in connection with the etiology of sarcoidosis as an allergic endogenous infection caused by this indigenous bacterium.

Results

Monophyly of the C1 sarcoid isolate in core genome analysis

The genomic sequence data for 77 strains of P. acnes were available from the database at the time of writing; the C1 strain of P. acnes is the only clinical isolate from sarcoid tissue for which the whole genomic sequence has been determined16. We first compared amino acid sequences of CDSs among all strains for which genomic sequences were available by sequence similarity.

A total of 1477 single-copy core CDSs were identified and 1262 of the 1477 CDSs were used for construction of a phylogenetic tree. In the maximum likelihood-based phylogenetic tree, the C1 strain was separately located as a monophyletic clade (Fig. 1), although only the C1 strain was included in the analysis due to unavailability of any other genome information of the isolates from sarcoid tissue samples (sarcoid isolates). These findings suggest that sarcoid isolates have evolved to be monophyletic.

Figure 1
figure 1

A maximum likelihood-based phylogenetic tree of 77 P. acnes constructed by 1262 core CDSs.

The tree was constructed by concatenated amino acid sequences of 1262 core CDSs among 77 P. acnes genomes. Detailed tree structure of a dense-branching part is shown in the upper box. Intricate parts in the main and detailed trees are indicated in grey and the strains in each part are shown altogether without precise location at the tree. Only bootstrap probability values over 70% are given. The sarcoidosis-derived strain is indicated by red color.

Full size image

Unique region on the genome of the C1 sarcoid isolate

Next, we compared whole genome sequences of the C1 sarcoid isolate with those from nine other strains available in the NCBI GenBank database using multiple genome alignment (Fig. 2). Homology was observed along whole genome in all dot plots, except inversions in C1 versus ATCC 11828 and C1 versus HL096PA1, which were reported previously17 (Fig. 2-i). All the breakpoints of these inversions were located in rRNA-encoding regions and these inversions seemed to occur symmetrically across the replication axes of the genomes. Moreover, the C1 genome contained an 18.8-kbp specific region that was absent in the other nine genomes (Fig. 2-ii). Disruption by this C1-specific region was observed in an alpha/beta hydrolase-encoding CDS, which was intact in the genomes other than C1. At both ends of this region, transposase-encoding CDSs were located between the two similar inverted-repeat sequences; the region was likely a composite transposon composed of two ISs (including transposase between two repeat sequences) at both ends and their intermediate CDSs between the two ISs (Fig. 2-iii).

Figure 2
figure 2

Multiple genome alignment of 10 P. acnes complete genomes.

(i) Dot plots of C1 against the nine other genomes are shown. Each dot indicates 20-bp match between two genomes and only ≥ 65-bp continuous dots are shown as a line. (ii) Multiple alignments of 10 P. acnes genomes are shown as a figure constructed by Mauve. Each colored box indicates a local collinear block (LCB) that is defined as a genomic region free from genome rearrangements and LCBs with the same color are linked by lines, indicating homology with each other. An 18.8-kbp C1-specific region is indicated by a dashed line box. (iii) A layout of CDSs in the C1-specific 18.8-kbp region and adjacent region is shown with the corresponding loci on 266 genome. Each boxed arrow indicates a CDS and the arrowhead is pointed in the transcriptional direction. Homology is indicated in grey, while the novel IS and putative transposon are indicated in yellow and green, respectively. In the transposon, the arrows are colored as follows: red, identical or homologous to the CDS of Propionibacterium humerusii; blue, identical or homologous to the CDS of Propionibacterium sp. 5U42AFAA strain; white, identical or homologous to the CDS of dairy propionibacteria. Amplicon sites from PCR with the primers in Supplementary Table S3 are indicated by bold lines.

Full size image

The ISs of the identified transposon were of a novel family because the sequences were not found in the public database. Max score and e-value of the most similar IS was 44.1 bits and 3e-04, respectively, based on the software ISfinder18. The identified transposon contained 13 hypothetical and 14 functionally-known CDSs; the latter included resolvase-coding and arsenate-related CDSs, such as those encoding arsenic resistance protein and arsenite-activated ATPase (Supplementary Table S1).

All 27 CDSs in the transposon were identical to the CDSs of species other than P. acnes, such as P. humerusii, P. jensenii, P. freudenreichii and P. acidipropionici.

Phylogenetic dispersiveness of sarcoid and non-sarcoid isolates in MLST analysis

MLST analysis was performed with 24 sarcoid and 36 non-sarcoid isolates, together with the reference ST data (ST1-ST93) available from the public database. The 76 P. acnes strains for which either complete or draft genome sequences were available in the public database were excluded in the MLST analysis, because their STs were already known and were therefore less informative (Supplementary Table S2). In a phylogenetic tree constructed from concatenated nucleotide sequences of 9 loci, 28 sequence types (STs) were identified among 60 isolates examined, including novel STs (ST94-ST112) (Table 1 and Fig. 3-i). ST26 isolates were most frequently found in 6 (25%) of 24 sarcoid isolates and 7 (19%) of 36 non-sarcoid isolates without a significant difference between them. The remaining (75%) sarcoid isolates were located dispersively across various STs. The dispersiveness of the sarcoid isolates was supported by differences in the allele number combination shown in the eBURST diagram, although STs of the sarcoid isolates were limited in number (Fig. 3-ii).

Table 1 Genetic profiles of 60 P. acnes strains by MLST and PCR analysis
Full size table
Figure 3
figure 3

A neighbor joining-based phylogenetic tree and an allelic profile diagram of 69 P. acnes strains and 93 reference STs.

(i) The tree was constructed by concatenated nucleotide sequences of nine loci in the P. acnes MLST. Detailed tree structure of a dense-branching part is shown in the upper box. The ST types are shown with the isolate names that were classified in the corresponding STs. Intricate parts in the main and detailed trees are indicated in grey and the strains in each part are shown altogether without precise location in the tree. Only bootstrap probability values over 70% are given. Isolates exhibiting cell-invasiveness are indicated in red. (ii) Diagram constructed by eBURST. Each circle indicates an allelic profile in the P. acnes MLST, with the ST number. The ST numbers including sarcoid isolates are indicated by open circles and the others are indicated by filled circles. The STs are single-locus variants against each other if they are linked by a line and singletons if not linked.

Full size image

ST26 isolates with the novel IS and four CDSs in the transposon

PCR detection of the novel IS was successful in 14 of the 60 isolates including the C1 strain (Table 1). All 13 ST26 isolates and a single ST91 isolate carried the novel IS. Most (12 of 13) of the ST26 isolates carried the four representative CDSs (hypothetical 15.9 kDa protein, arsenic resistance protein, regulatory protein ArsR and resolvase) that were contained in the unique transposon sequence, with one exceptional strain in which the IS was positive but the four genes were totally negative based on PCR.

ST26 isolates with cell-invasiveness

Comparison of the genomic profiles of P. acnes examined in the present study with the cell-invasiveness of each strain reported in our preceding study19 revealed that 12 (50%) of the sarcoid isolates and 16 (44%) of the non-sarcoid isolates were cell-invasive (Supplementary Table S2). Cell-invasive strains were classified in a limited number of STs (ST8, 26, 36, 41, 67, 70, 100, 112) among a total of 28 STs found in all isolates (Fig. 3-i). All of the ST26 isolates with the novel IS were cell-invasive.

Discussion

We previously reported a whole genome sequence of the C1 strain of P. acnes from a granulomatous inflammatory lesion of a sarcoidosis patient16. To search for a specific genetic profile of this sarcoid isolate, we first performed core genome analysis with whole genome sequences from 76 P. acnes strains and multiple genome alignment with complete genome sequences from 9 P. acnes strains. The genomic profiles we found in this sarcoid isolate led to the identification of a transposon unique to the C1 isolate with a novel IS. P. acnes strains with the novel IS were classified in ST26 by MLST, with one exception (ST91). PCR analysis for 4 CDSs of the transposon suggested that most of the P. acnes strains with the novel IS carry the transposon, which may allow us to determine relevant factors of the bacterium in the etiology of sarcoidosis.

In the present study, ST26 was phylogenetically independent from the other STs based on the core genome analysis (Fig. 1). In the MLST analysis, the ST26 and ST91 strains were phylogenetically independent (Fig. 3-i); however, only ST26 and not ST91, exhibited cell-invasiveness, indicating the phylogenetic independence of ST26 from the others, which was not apparent in the MLST analysis due to the use of only house-keeping CDSs and lack of sufficient genetic information. According to the results by Lomholt and Kilian12, ST26 of P. acnes is different from other ST groups of P. acnes in terms of the mutational status of the two hemolytic-associated genes (camp 5 and tly) of this indigenous bacterium. The present study demonstrated the phylogenetic independence of ST26 based on the core genome analysis delineating the P. acnes population with high resolution.

A well-known genotype of P. acnes is recA (types I, II and III). Each of the recA genotypes has a characteristic phenotype and recA type I is dominant in isolates from acne vulgaris20. Based on the genotype of the recA gene against the P. acnes isolates in this study, all the ST26 strains in this study were classified as type I. The isolates in type I are prevalent in acne vulgaris and exhibit beta-hemolysis21. The study of clustered regularly interspaced short palindromic repeats (CRISPR) in P. acnes revealed that CRISPR were present exclusively in types II and III and differentiated type I from type II20. Absence of the CRISPR in type I strains is consistent with the presence of the novel transposon in ST26 strains in type I. The ST26 strains might have evolved to be genetically and phenotypically unique in type I, which is the type possibly evolved from type II.

With regard to the uniqueness of ST26 P. acnes strains, it is notable that the presence of the unique transposon in the genome was suggested in most ST26 P. acnes strains, as well as in the C1 strain. This transposon carried not only functionally known CDSs, such as those encoding arsenical- and metal-resistance proteins, but also hypothetical CDSs for which cell-invasiveness of P. acnes seems to be essential for linking this indigenous bacterium to the cause of sarcoidosis, because infectious granulomas are commonly caused by intracellular pathogens. The cell-invasiveness of P. acnes is closely associated with the serotype and particular genotypes19. In the present study, cell-invasiveness was correlated with a limited number of STs among a total of 28 STs found in the isolates examined. Because all ST26 strains of P. acnes were cell-invasive, ST26 strains might have evolved to acquire advantageous characteristics for intracellular persistence of the bacterium after cell-invasion by unknown mechanisms, including horizontal gene transfer via transposition of particular genes. The previous study suggested that specificity of genetic elements to each P. acnes lineage contributes to phenotypic and functional differences of P. acnes as a commensal and pathogenic agent22,23. Considering that a plasmid found in a P. acnes strain is suggested to be associated with P. acnes virulence17, the novel transposon might confer novel genetic characteristics to the strains of this unique ST.

The lack of a genetic profile specific to the sarcoid isolates, however, has been reported in several studies. Ishige et al.15 reported that P. acnes isolates were not specific to sarcoidosis when examined by random amplified polymorphic DNA analysis with 45 sarcoid and 67 non-sarcoid isolates. Furukawa et al.19 also reported that P. acnes isolates were not specific to sarcoidosis in terms of serotype, cell-invasiveness, or genetic polymorphism of the trigger factor gene and the two invasion-associated P. acnes genes. Based on the lack of any specific characteristic of the sarcoid isolates, they concluded that host factors that cause an allergic Th1 immune response to the indigenous bacterium are more important for the onset of sarcoidosis than pathogen factors. The present study, however, demonstrated that the sarcoid isolates were likely to have evolved uniquely; the sarcoid isolates might have the capacity to induce chronic inflammation and unknown factors carried by the transposon unique to the ST26 isolates might be associated with such a characteristic of the sarcoid isolates.

Also, in the present study, ST26 of P. acnes with the novel transposon was not specific to sarcoid isolates. The lack of P. acnes strains specific to sarcoidosis does not exclude the possibility that a certain strain of P. acnes causes sarcoidosis in a genetically susceptible subject under certain environmental conditions. A single isolate from each sarcoid sample does not always represent the P. acnes strain that causes sarcoid lesions due to heterogeneity of characteristics in the population. Most of the sarcoid isolates were cultured from lymph nodes affected by sarcoidosis. This indigenous bacterium is also isolated from some non-sarcoid lymph node samples. Such non-pathogenic strains cannot be discriminated from pathogenic strains when a single colony is picked up from a culture plate as a representative isolate from the sarcoid sample. The C1 strain is an exceptional sarcoid isolate cultured from a sarcoid granulomatous inflammatory lesion in the subcutaneous fatty tissue. Because P. acnes has never been found in non-sarcoid subcutaneous tissue, it is free from indigenous flora and seems to be isolated only from the sarcoid granulomatous-inflammatory lesions.

In conclusion, we demonstrated the phylogenetic independency of ST26 strains and their unique characteristics of cell-invasiveness and a unique transposon and suggested that ST26 is a responsible agent for sarcoidosis. Further studies of ST26 such as whole genome analysis of ST26 P. acnes isolates other than C1 are essential for elucidating possible pathogenic factors of this indigenous bacterium in the etiology of sarcoidosis.

Materials and Methods

P. acnes strains

A total of 60 P. acnes isolates were evaluated (Supplementary Table S2). All of the P. acnes isolates used for the study were collected earlier15,19. A representative strain (C1) of sarcoid isolates, which was used for the previous complete genome sequence analysis by Minegishi et al.16, was isolated from a subcutaneous lesion of a 25-year-old woman with sarcoidosis. Of the other 59 P. acnes isolates evaluated, 23 were isolated from 23 lymph nodes of 23 patients with sarcoidosis, 10 were isolated from 10 non-metastatic lymph nodes draining from the stomach, lung, or colon with primary cancer (4, 3, 3 strains, respectively), 12 were isolated from skin swabs of 12 healthy individuals and 14 were isolated from prostate tissue of 14 patients with prostate cancer. Genomic information of 76 strains (9 complete and 67 draft genomes) was available from the DDBJ/EMBL/GenBank database.

Culture condition and DNA extraction

Stored isolates of P. acnes were grown in Gifu anaerobic medium (GAM) broth (Nissui Pharmaceutical Co., Ltd., Tokyo, Japan) at 37°C under anaerobic conditions (10% H2, 10% CO2, 80% N2) for 3 days. Isolation of genomic DNA was described previously19.

Core genome analysis

All 77 P. acnes genome sequences (see above section “P. acnes strains”; C1 genome and 76 genomic information in the public database) were processed by the RAST server24,25 for prediction of CDS regions with functional annotation and used for the following analysis as the information derived under the same CDS-prediction/annotation criteria. The amino acid sequences of all the predicted CDSs were clustered by PGAP v1.02 under the default parameters26. Single-copy core CDSs were identified as those that were located in a single genomic region and commonly present on all the genomes, while strain-specific CDSs were identified as those found exclusively on a single genome. The amino acid sequences of the single-copy core CDSs were concatenated in each strain after exclusion of the CDSs with endogenous rearrangement events by a Phi test, which is a partial algorithm of SpritsTree427,28. The concatenated amino acid sequences were used for construction of a maximum likelihood-based phylogenetic tree. ModelGenerator v851 was used to estimate the appropriate substitution model of amino acid and RAxML v7.2.8 was used for tree construction under the Jones-Taylor-Thornton model and 100 times bootstrap iteration29,30. The tree was visualized by Dendroscope v331,32.

Multiple genome alignment

The complete genome sequences of 10 P. acnes strains were aligned using the nucmer program in MUMmer v3.23 and progressiveMauve mode in Mauve v2.3.133,34. The C1-specific region was identified as a gap in the Mauve alignment and ISsaga was used to identify any ISs in the C1-specific region35. The identified IS was considered to be novel if all alignments between the identified IS and each of any known ISs in the ISsaga database had >80% length of the known IS and >80% nucleotide identity.

Annotation of the CDSs in the unique transposon

Annotation of the CDSs in the unique transposon was based on the results of BLASTP searches against the NCBI nonredundant protein database36,37.

PCR conditions

The novel IS and several intermediate CDSs in the transposon were detected by PCR in 59 P. acnes isolates using the primers listed in Supplementary Table S3. The PCR conditions for the IS were as follows: 5 min at 94°C, followed by 30 cycles of 30 s at 94°C, 30 s at 58°C and 80 s at 72°C. The PCR conditions for the other genes were 3 min at 94°C, followed by 30 cycles of 30 s at 94°C, 30 s at 58°C and 90 s at 72°C. The annealing temperature for arsenic resistance protein was 60°C. The PCR was completed with a final extension step at 72°C for 7 min. Location of the amplicons on C1/266 genomes is shown in Fig. 2.

MLST analyses

We used nine genetic loci (cel, coa, fba, gms, lac, oxc, pak, recA and zno) for the MLST analyses. PCR conditions and characterization of the allelic profiles were described previously12,38. The nucleotide sequences of all nine loci were concatenated to use for the construction of a neighbor joining-based phylogenetic tree by MEGA v5.2 under Kimura’s two-parameter (K2P) substitution model and 1000 times bootstrap iteration39. The allelic profiles were visualized by drawing a diagram using eBURST v340. The sequence data and allelic/ST profiles available in the public database (http://pacnes.mlstransposonet) were included in the above MLST analyses.

Nucleotide sequence accession numbers

Nucleotide sequences of the MLST analyses have been deposited in the DDBJ/EMBL/GenBank databases under the following accession numbers: LC006312-LC006851.