Comparative Screening of Digestion Tract Toxic Genes in Proteus mirabilis.

Proteus mirabilis is a common urinary tract pathogen, and may induce various inflammation symptoms. Its notorious ability to resist multiple antibiotics and to form urinary tract stones makes its treatment a long and painful process, which is further challenged by the frequent horizontal gene transferring events in P. mirabilis genomes. Three strains of P. mirabilis C02011/C04010/C04013 were isolated from a local outbreak of a food poisoning event in Shenzhen, China. Our hypothesis is that new genes may have been acquired horizontally to exert the digestion tract infection and toxicity. The functional characterization of these three genomes shows that each of them independently acquired dozens of virulent genes horizontally from the other microbial genomes. The representative strain C02011 induces the symptoms of both vomit and diarrhea, and has recently acquired a complete type IV secretion system and digestion tract toxic genes from the other bacteria.

Introduction

The gram-negative anaerobic bacterium Proteus mirabilis is one of the major Proteus infectious factors, and may cause severe pain by forming stones in the urinary tract [1]. P. mirabilis may induce various pathogenic symptoms, including fever, chills and chest pain, etc [2]. Its motility and adhesion to solid surface makes P. mirabilis extremely easy to propagate through medical devices in hospitals [2]. P. mirabilis is also notorious for its ability to actively acquire antibiotic resistance and infectious toxin genes from other microbial genomes through horizontal gene transferring [3, 4].

Quite a number of molecular typing technologies have been developed for a prompt screening of Proteus mirabilis in clinical samples. The protein UreC is the alpha subunit of the urea degradation pathway encoded by P. mirabilis [5]. The second metabolite repressor gene rsmA is involved in the swarming motility and the expression regulation of virulence factors in P. mirabilis [6]. P. mirabilis also encodes a kidney damage inducer, the hemolysin HpmA, which shows significant cytotoxicity [7]. The virulence factor metalloprotease ZapA is demonstrated to regulate the IgA hydrolysis process by the inhibiting profiles [8]. A number of social network analyzing techniques were applied to characterize the disease-associated microRNAs [9, 10], and the improved detection performance suggests that techniques from different areas may be complementary to each other for the challenging detection problem for the disease-microRNA association [11, 12], which may benefit the detection problem of pathogenicity genes in P. mirabilis.

This study hypothesized that P. mirabilis exerts its infectious and toxic functions in the new location, digestion tract, through site specific genes. Three P. mirabilis strains isolated from a local outbreak of food poison, and two more non-toxic strains were isolated from the same physical location after the outbreak event. The genes and repetitive elements were annotated for all the P. mirabilis strains. A comprehensive screening of the digestion tract toxicity (DTT) specific genes is carried out for the P. mirabilis infectious strains. The comparative study supports our hypothesis that some strain specific genes may serve as candidates for rapid strain typing and treatment selection.

Material and Methods

Strain isolations

During an outbreak of a food-poisoning event in Shenzhen in June of 2002, three strains of P. mirabilis were collected from samples of different patients. The strain C02011 was isolated from the vomit sample of a patient, and it has the same Pulsed-Field Gel Electrophoresis (PFGE) band of the P. mirabilis strain isolated from the diarrhea stool. So the strain C02011 is kept as a representative strain for further analysis. The strain C04010 was isolated from the diarrhea stool sample of a patient in the same event. Another strain C04013 was isolated from the food taken by the patients immediately before this food-poisoning event.

These three strains of P. mirabilis were compared with two locally isolated strains and two reference strains. Another two strains of P. mirabilis C02034 and B02005 were isolated from the food and food worker’s stool during non-outbreak time at the same physical location. The two reference strains HI4320 [2] and BB2000 [13] were urinary tract infection strains, and were also chosen as the control samples for the comparative genomics investigation.

This study was approved by the Ethics Committee of Shenzhen Center for Disease Control and Prevention on June 9, 2002. The authors did not collect the samples directly from the patients, and received the samples from the local hospitals. This study investigated the genome sequencing data of the pathogens and did not access the patients’ personal data. So no informed consent forms were collected from the patients.

Genomic sequencing and gene annotation

The five newly isolated strains were sequenced and assembled by the BGI-Shenzhen. Two insert fragment lengths 500 and 2000 bps were chosen for the paired-end (PE) sequencing using the Illumina HiSeq-2000 machine. The read lengths were set as PE-90 and PE-49 for the 500-bp and 2000-bp libraries, respectively. The 500-bp paired-end reads were de novo assembled using the program SOAPdenovo version 1.06, and the contigs were further connected into scaffolds using both 500-bp and 2000-bp PE reads [14]. The genes were annotated using the program Glimmer version 3.02 [15]. The functions of each gene were annotated using the NCBI BLAST based on the databases NCBI NR [16], SwissProt/Trembl [17], COG [18] and KEGG [19], respectively. The data are summarized in the Results and discussion section. Transcriptome is not investigated in this study and the microRNA precursor genes will be characterized by both in silico techniques [20] and RNA-seq experiments [21].

Annotation of repetitive elements

Repetitive elements are not annotated by default for a genome sequencing service, and a comprehensive annotation is conducted for the five sequenced strains and the two reference strains of P. mirabilis investigated in this study.

Repetitive elements were annotated by the RepeatMasker version 4.05 [22]. The tandem repeats were screened by the de novo program TRF version 4.04 [23]. Prokaryotic genomes have transposable elements like Insertion Sequence (IS) elements [24, 25] and Miniature Inverted-repeat Transposable Elements (MITEs) [26-28]. Full copies of IS elements were detected and curated by multiple sequence alignments of IS encoded transposase genes, using the program MEGA version 6.06 [29]. Full copies of detected IS elements and all other full-length IS copies from the database ISfinder [30] were used as templates to screen for all the IS copies in the seven strains of P. mirabilis. No MITEs were detected by the program MUST version 1.0 [31].

CRISPRs are a series of consecutive short repeats with spacers in between, and the spacer sequences are usually acquired from the invasive foreign elements [32, 33]. The program CRISPRfinder [34] was used to detect CRISPRs in the seven investigated P. mirabilis genomes.

Generation of the phylogenetic tree

A phylogenetic tree based on the 16S ribosomal RNA (rRNA) genes is generated for the five P. mirabilis strains sequenced by this study and the three additional P. mirabilis reference strains published elsewhere. The 16S rRNA genes of the five newly sequenced P. mirabilis strains are annotated as described in the above sections. The 16S rRNA genes of the two P. mirabilis strains HI4320 and BB2000 are collected from the NCBI database. C05028 is another publicly available strain with DTT function [35], and its 16S rRNA gene is extracted using NCBI MegaBlast version 2.2.32+ [36] with the 16S rRNA of P. mirabilis HI4320 as template. The phylogenetic tree is rooted at Proteus penneri ATCC 35198, whose 16S rRNA gene is detected with the template gene from the strain HI4320 using the program NCBI BLASTN version 2.2.32+ [36]. The E-value 0.0 supports that the two microbial genomes are close relatives with enough statistical significance. The phylogenetic tree is generated using Phylogeny.fr [37]. All the programs are executed using the default parameters.

Screening for the strain specific genes

A highly sensitive screening of whether a gene has a homolog in a given genome is conducted using the NCBI MegaBLAST version 2.2.32+ [36] with the default parameters. Only those genes with no homologs are kept as strain specific candidates.

Results and Discussion

In vitro PFGE strain typing

Three isolated P. mirabilis strains demonstrate an abnormal infectious ability and toxicity to the digestion tract, but two other strains don’t have such ability. So the first question is whether the existing molecular typing technologies may differentiate these strains by their pathogenesis.

PFGE is a widely used rapid subtyping technique for the macrorestriction analysis of pathogens, as recommended by the PulseNet laboratories [38]. 50 U of the restriction enzyme SfiI in the 200 ml buffer was used to generate the genomic fragment profile, and a standard protocol was followed [39]. After the gel was scanned as the TIFF format image, a dendrogram tree was calculated using the software BioNumerics version 5.0 (Applied Maths BVBA, Belgium), as similar in [40].

The restriction enzyme SfiI cuts the genomic sequence into fragments of lengths between 20–700 kbps, and the PFGE profiles in Fig 1 shows significant differences even between closely related P. mirabilis strains. The inconsistency of PFGE profiles of these six P. mirabilis strains with their pathogenicity suggests that these genomes may under strong selection pressures for genomic changes, and may undergo both intra-genome structural variations and inter-genome/strain horizontal transferring.

Fig 1

The dendrogram tree of six P. mirabilis strains.

Strains C05028 and C02011 are digestion tract infectious. Strains HI4320 and ATCC29906 are urinary tract infectious. Strains B02005 and C02034 are non-infectious strains.

In vitro PCR strain typing

The polymerase chain reaction (PCR) is also a common technique to subtype pathogens based on the existences of known pathogenic factors [41]. The PCR-based screening of the genes ureC, rsmA, hpmA and zapA were conducted in the five newly isolated strains of P. mirabilis. As shown in Table 1, the primers for these four genes were designed in-house, and a standard PCR protocol was followed.

Table 1

PCR primers for the four genes.

Target	pName	Primer (5’-3’)	Product (bps)
ureC	ureC-F	GTATTCGTCGCGAAACCATT	553
	ureC-R	ATATGACGACAGCCCTCCAC
rsmA	rsmA-F	TAGCGAGTGTTGACGAGTGG	562
	rsmA-R	AGCGAGGTGAAGAACGAGAA
hpmA	hpmA-F	CCAGTGAATTAACGGCAGGT	654
	hpmA-R	CGTGCCCAGTAATGGCTAAT
zapA	zapA-F	AACGCAGGTCAGAATGTTCC	553
	zapA-R	TATCTTGTCCACGACCACCA

Column pName lists the names of the primers.

Unfortunately, there is no difference between the digestion tract toxic strains and the other strains, since all the five strains encode all these four genes, as shown in Table 2 and S1 Fig. So the existing pathogen subtyping techniques are not precise enough to differentiate these five P. mirabilis strains.

Table 2

Currently adopted molecular typing technologies of biomarker genes in the infectious bacterial strains.

Strain	Detail	ureC	rsmA	hemolysin (hpmA)	metalloprotease (zapA)
C02011	Patient vomit	+	+	+	+
C04010	Patient stool	+	+	+	+
C04013	Poison food	+	+	+	+
C02034	Normal food	+	+	+	+
B02005	Normal stool	+	+	+	+

The first three strains C02011, C04010 and C04013 induce DTT function, while the other two strains don’t.

This study carried out a comprehensive screening of functional elements that are specific to these three digestion tract toxic strains, compared against the other two digestion tract infectious strains. These strain specific genes may serve as candidate subtyping targets.

Summary of the seven P. mirabilis strains

P. mirabilis strain C02011 induces the symptoms of both vomit and diarrhea, and is chosen as the reference genome for the screening of DTT genes. Among the seven P. mirabilis strains, C02034 and HI4320 have the genome sizes larger than 4 Mbps, and all the other strains have a genome size between 3.80 and 3.85 Mbps, as shown in Table 3. It’s interesting to observe that these two largest genomes also have the largest numbers of IS elements, with 28 and 24 in C02034 and HI4320, respectively. This supports the observation that IS number tends to be positively correlated with the genome size [25, 42]. All the seven strains have a similar GC content between 38.38–38.88%. The Pearson’s test generates the P-value = 7.341e-4 for the correlation between the Gsize and Gene# for the seven strains of P. mirabilis. So a large genome tends to encode more genes. Only the two strains C02011 and C02034 have 2 CRISPRs, respectively.

Table 3

Annotation summary of the seven P. mirabilis strains investigated in this study.

Strain	Gsize (bps)	GC (%)	Gene#	IS	CRISPR	rDrug	Toxin	T4SS	DTT specific
C02011	3802255	38.45	3,463	20	2	45	17	14	62
C04010	3803688	38.38	3,465	17	0	46	30	3	66
C04013	3844051	38.46	3,511	11	0	46	10	3	59
B02005	3830296	38.56	3,516	12	0	45	0	3	-
C02034	4166171	38.62	3,761	28	2	51	0	3	-
BB2000	3846754	38.60	3,455	15	0	42	-	3	-
HI4320	4063606	38.88	3,607	24	0	41	-	3	-

Columns “Gsize (bps)” and “GC” give the estimated genome size and the G+C content of each strain. The numbers of annotated genes and IS elements are given in the columns “Gene#” and “IS”. The annotated CRISPR number is given in the column “CRISPR”. The numbers of drug resistant genes and toxin genes are based on the gene function annotation and listed in the columns “rDrug” and “Toxin”, respectively. The gene number in the T4SS is in the column “T4SS”, which is summarized from the gene annotations. The column “DTT specific” gives the number of genes with no homologs in the non-toxin P. mirabilis strains. “-”means that this item is not analyzed for this strain. The gene annotation of the strain C05028 is not publicly available, and is not included in this table.

All the seven P. mirabilis strains encode drug resistance genes, but only the three strains with DTT function encode toxin genes, as shown in Table 3. The type IV secretion system (T4SS) does not seem to be a major secretion system in P. mirabilis, with 3 T4SS genes in each strain except for C02011. And each of the three digestion tract toxic strains encodes ~60 genes, which are not detected in the other P. mirabilis strains.

This study will focus on the DTT specific genes that may facilitate the radical pathogenesis during the local food poison event.

Phylogenetic traces

P. mirabilis is notorious for its infection and stone forming in the urinary tract, and usually causes severe pain. The two publicly available strains BB2000 and HI4320 may induce various whole body symptoms when infecting the urinary tract [2, 13]. But the five strains isolated from Shenzhen and the strain C05028 seem to have gained the ability to infect the digestion tract, and four of them demonstrate DTT function during their infections.

The five locally isolated strains of P. mirabilis together with the strain C05028 originate from the same common ancestor, whereas the other two urinary tract infectious strains BB2000 and HI4320 originate from another common ancestor, based on the phylogenetic tree in Fig 2. The two non-toxic strains C02034 and B02005 do not show a clear separation from the three digestion tract toxic (DTT) strains on the phylogenetic tree. This suggests that genes specific to DTT strains may play an essential role in infecting and inducing toxicity to the digestion tract.

Fig 2

The phylogenetic tree of the seven P. mirabilis strains based on the 16S rRNA genes.

The Cladogram tree is rooted at Proteus penneri ATCC 35198, and generated by the Phylogeny.fr. The three P. mirabilis strains inducing DTT function are highlighted in the bold font.

DTT specific elements

Most of the DTT specific genes are also strain specific, as shown in Fig 3. The DTT specific genes have no homologs in the two non-toxic strains C02034 and B02005 and the two urinary tract infectious strains HI4320 and BB2000, and over 70% of them are only detected in one of the three DTT strains. And all 45 genes specific in the strain C02011 does not have detectable homologs in the publicly available strain P. mirabilis C05028. By considering the phylogenetic relationships among the three DTT strains (as in Fig 2), the current data strongly suggests that all the three P. mirabilis DTT strains are actively and independently acquiring foreign genes from the environmental neighboring microbes.

Fig 3

Venn plot of the strain specific genes in the three P. mirabilis DTT strains.

The numbers in the brackets are the numbers of DTT specific genes compared against the other P. mirabilis strains.

Core module for DTT function

The most infectious and toxic strain C02011 encodes 45 genes with no homologs in all the seven other strains. This suggests that these genes may have been acquired through the mechanism of horizontal gene transferring. The function annotation shows that these 45 genes do not introduce new antibiotic resisting abilities into the strain C02011, as shown in Table 4 and S1 Table.

Table 4

45 genes specific in P. mirabilis C02011.

Scaffold	Start	End	Strand	gene_id	Annotated protein
fa_scaffold3	465840	465962	+	C02011GL001038
fa_scaffold3	465952	466815	+	C02011GL001039
fa_scaffold3	466833	467849	+	C02011GL001040	Putative phage gene
fa_scaffold3	467846	468379	+	C02011GL001041
fa_scaffold3	468467	468952	+	C02011GL001042	Putative phage gene
fa_scaffold3	468949	469179	+	C02011GL001043
fa_scaffold3	469224	471878	+	C02011GL001044	Putative inner membrane protein
fa_scaffold3	472166	472393	+	C02011GL001045	Putative phage gene
fa_scaffold3	472383	472538	+	C02011GL001046
fa_scaffold3	473119	473679	+	C02011GL001047
fa_scaffold3	473682	475016	+	C02011GL001048	Retro-type reverse transcriptase
fa_scaffold3	1068445	1069563	-	C02011GL001596
fa_scaffold3	1069568	1070161	-	C02011GL001597
fa_scaffold3	1535544	1536122	-	C02011GL002036
fa_scaffold3	1536545	1537150	+	C02011GL002037	Integrase
fa_scaffold3	1538064	1538282	+	C02011GL002039
fa_scaffold3	1538847	1539260	-	C02011GL002041
fa_scaffold3	1543281	1543748	-	C02011GL002049
fa_scaffold3	1543975	1544586	-	C02011GL002050
fa_scaffold3	1544614	1545003	-	C02011GL002051
fa_scaffold3	1809941	1810177	-	C02011GL002280
fa_scaffold3	1810202	1810420	-	C02011GL002281
fa_scaffold3	1810632	1811384	-	C02011GL002282	Putative phage regulatory protein
fa_scaffold3	1811504	1812076	-	C02011GL002283	BRO family protein
fa_scaffold3	1812403	1813122	+	C02011GL002284
fa_scaffold3	1823562	1824128	-	C02011GL002300	Putative phage terminase
fa_scaffold3	1824196	1824423	+	C02011GL002301
fa_scaffold3	1824634	1824879	-	C02011GL002302
fa_scaffold3	1825050	1825649	-	C02011GL002303	KilA-N domain-containing protein
fa_scaffold6	547377	548129	-	C02011GL003316	MobC
fa_scaffold6	548138	549889	-	C02011GL003317	MobB
fa_scaffold6	549859	550503	-	C02011GL003318
fa_scaffold6	550993	551298	-	C02011GL003319	ORF15
fa_scaffold6	551334	551708	-	C02011GL003320
fa_scaffold6	551729	552754	-	C02011GL003321	T4SS virB11
fa_scaffold6	552744	554012	-	C02011GL003322	T4SS virB10
fa_scaffold6	554056	554961	-	C02011GL003323	T4SS virB9
fa_scaffold6	554961	555644	-	C02011GL003324	T4SS virB8
fa_scaffold6	555866	556936	-	C02011GL003325	T4SS virB6
fa_scaffold6	557184	557885	-	C02011GL003326	T4SS virB5
fa_scaffold6	557903	560641	-	C02011GL003327	T4SS virB4
fa_scaffold6	560654	560833	-	C02011GL003328	T4SS virB2
fa_scaffold6	560947	561699	-	C02011GL003329	T4SS virB1
fa_scaffold6	562549	563136	-	C02011GL003330
fa_scaffold6	590486	591754	-	C02011GL003343	Integrase

The three large genomic islands are highlighted by a bold font in the column “gene_id”. The gene annotations are summarized from the S1 Table.

Three large genomic islands with at least 5 consecutively distributed genes are detected in the strain C02011. The largest genomic island encodes 15 genes, including a complete T4SS. The transmembrane secretion assembly T4SS consists of multiple component proteins, and is utilized by various gram-negative bacteria to transport virulent factors between the pathogen and the host cells [43]. Among the sequenced P. mirabilis strains, T4SS is only observed in the urinary tract infecting strain HI4320 [2, 44, 45]. And the T4SS in the strain C02011 does not show any similarity on the primary sequence level to that in HI4320, which supports the observation of little sequence conservation of T4SS modules from different bacteria [43, 46–49]. This genomic island also encodes three genes with significant homologs in the plasmid pMET-1 from the Gammaproteobacterium Klebsiella pneumoniae[50]. Among the three genes, mobB encodes three transmembrane domains and ORF15 has a 5’ signal peptide.

Another two genomic islands encode genes with various phage functionalities, as shown in the S1 Table.

Besides these transmembrane transporting genes, the strain C02011 also encodes a gene C02011GL002300 homologous to the phage terminase from a food poison pathogen Yersinia enterocolitica 8081, as shown in the S1 Table. Y. enterocolitica 8081 belongs to the serotype 0:8 and biotype 1B of Y. enterocolitica, and may induce severe diarrhea [51]. The most recent common ancestor between P. mirabilis and Y. enterocolitica on the phylogenetic tree is Enterobacteriaceae on the family level. The data suggests that P. mirabilis C02011 may have horizontally acquired severe digestion tract toxic genes from another genus Yersinia.

PCR products after gel electrophoresis of the four genes in the five investigated strains of P. mirabilis.

(a) ureC, (b) rsmA, (c) hpmA, and (d) zapA. The lanes M and C are the 100-bp ladder and the control ddH2O lanes, respectively. The lanes 1 through 5 are the P. mirabilis strains C02011, C04010, C04013, C02034 and B02005, respectively.

(DOCX)

Click here for additional data file.

Annotations of the 45 genes specific in P. mirabilis C02011.

The three candidate genomic islands are highlighted by bold font in the column “gene_id”. The annotation is provided by the BGI, using the reference databases NCBI NR, SwissProt, Trembl, COG and KEGG.

(DOCX)

Click here for additional data file.