Determination of the Relationship between Group A Streptococcal Genome Content, M Type, and Toxic Shock Syndrome by a Mixed Genome Microarray

Abstract

Group A streptococci (GAS), or Streptococcus pyogenes, are associated with a remarkable variety of diseases, ranging from superficial infections to life-threatening diseases such as toxic-shock-like syndrome (TSS). GAS strains belonging to M types M1 and M3 are associated with TSS. This study aims to obtain insight into the gene profiles underlying different M types and disease manifestations. Genomic differences between 76 clinically well characterized GAS strains collected in The Netherlands were examined using a mixed-genome microarray. Inter-M-type genomic differences clearly outweighed intra-M-type genome variation. Phages were major contributors to observed genome diversification. We identified four novel genes, including two genes encoding fibronectin-binding-like proteins, which are highly specific to a subset of M types and thus may contribute to M-type-associated disease manifestations. All M12 strains were characterized by the unique absence of the citrate lyase complex and reduced growth under hypoxic, nutrient-deprived conditions. Furthermore, six virulence factors, including genes encoding a complement-inhibiting protein (sic), an exotoxin (speA), iron(III) binding factor, collagen binding factor (cpa), and fibrinogen binding factor (prt2-like), were unique to M1 and/or M3 strains. These virulence factors may contribute to the potential of these strains to cause TSS. Finally, in contrast to M-type-specific virulence profiles, we did not identify a common virulence profile among strains associated with TSS irrespective of their M type.

Group A streptococci (GAS), or Streptococcus pyogenes, are human pathogens able to cause diseases ranging from superficial lesions to rapidly progressing and often fatal conditions such as necrotizing fasciitis and toxic-shock-like syndrome (TSS). Despite the persisting sensitivity of GAS to penicillin and other antibiotics, a resurgence of GAS infections has been reported since the mid-1980s (12, 19, 34, 46).

GAS are genotyped on the basis of the emm gene, encoding the M protein. Although all M types can give rise to severe GAS disease manifestations, the M1 and M3 types are overrepresented in the severest complication of invasive GAS disease, i.e., TSS. In our Dutch surveillance system for invasive GAS disease, M1 and M3 isolates have been associated with TSS in 41% and 39% of cases, respectively, whereas other M types have given rise to TSS in 23% of cases (56). Other specific M types are associated with pharyngitis and acute rheumatic fever or with skin infections and glomerulonephritis (13, 56). The M type bias in GAS disease manifestations is not absolute: within a given M type, strain-specific virulence characteristics also seem to contribute (25).

Several studies have been conducted to unravel the relationship between the genetic profiles of GAS strains of different M types and their clinical manifestations. Whereas M-type-specific gene profiles have been described, M-type-independent gene profiles associated with severe GAS disease have not been observed (45). One explanation is that, apart from M-type-related genetic differences, invasive GAS strains differ in gene regulation rather than gene profile (7, 54). Gene profiling studies, however, have largely been limited to a single M type or to virulence factors that have been identified previously. Although microarray analyses do provide information on the whole genome, their design is restricted by the sequenced GAS genomes available, which are predominantly from North American strains (9, 49). Therefore, an alternative explanation may be that not all relevant genes involved in invasive GAS disease have been uncovered yet. This study explores the latter hypothesis and aims to identify novel genes associated with invasive M types and TSS.

To circumvent the limitations mentioned above, we have developed a novel approach. In this approach, random DNA fragments obtained from different well-characterized GAS strains are used to produce a mixed-genome DNA microarray. The method does not require prior genome sequence information, allows genomic screening of a large strain collection, and enables the identification of new genes. This microarray was screened with a unique, clinically well documented GAS strain collection from The Netherlands (56, 57). Genetic differences may be most pronounced between strains representing the clinical extremes of GAS infections. Therefore, isolates associated with the severest manifestation of invasive GAS disease (TSS) were compared to isolates associated with mild, superficial infections.

First, we looked for a common GAS gene profile that correlates with TSS, independently of the M type. Subsequently, the genetic basis for the M type bias in TSS was explored. Genes common to the TSS-associated M types M1 and M3 were compared to the gene profiles of other M types. Finally, the main genetic differences between the GAS strains of different M types were determined. Understanding the gene profiles underlying different M types and TSS may deepen our understanding of GAS pathogenesis.

MATERIALS AND METHODS

Bacterial strains.

A total of 76 clinically well documented isolates representing the predominant M types M1 (n = 15), M3 (n = 14), M12 (n = 14), M28 (n = 14), M4 (n = 5), M6 (n = 5), M11 (n = 5), and M89 (n = 4) were selected from a Dutch surveillance program on GAS disease between 1992 and 2003 (56, 57). These M types are responsible for 74% of invasive GAS infections in The Netherlands (57). For each M type, isolates associated with TSS and those associated with noninvasive, superficial GAS disease were equally represented. For simplicity, we use the term “TSS strains” to designate strains isolated from normally sterile sites of patients with severe GAS infections that met the clinical criteria for TSS (59). Similarly, we use the term “superficial GAS disease strains” or “non-TSS strains” for strains isolated from patients with nonsevere, superficial GAS infections. Per M type, the distribution of TSS versus superficial GAS disease strains was as follows: for M1, eight versus seven isolates; for M3, seven versus seven; for M12, six versus eight; for M28, seven versus seven; for M4, three versus two; for M6, two versus three; for M11, three versus two; and for M89, two versus two. The M genotype was determined by hybridization of the denatured emm amplicon with a panel of 26 emm type-specific probes in a reverse line blotting system (29). The M genotype was confirmed by “conventional” emm typing for a subset of 16 strains (2 random strains per M type [data not shown]) (6). Strains were grown on blood agar plates under 5% CO₂ at 37°C overnight. Growth comparison experiments were conducted as follows. Strains were grown overnight in Todd-Hewitt broth supplemented with yeast extract (THY) under 5% CO₂ at 37°C. Overnight cultures were adjusted to an optical density at 660 nm (OD₆₆₀) of 1 with phosphate-buffered saline and diluted 1:10,000 in fresh, prewarmed THY medium. Anaerobic growth was measured every 20 min with an automated OD reader (Bioscreen C MBR System; Growth Curves AB Ltd).

PFGE typing.

A subset of 25 representative strains was subjected to pulsed-field gel electrophoresis (PFGE) as previously described (55). Briefly, the bacteria were suspended, and bacterial plugs were treated with lysis buffer for 48 h. The bacterial DNA was digested with the SmaI restriction endonuclease. Restriction fragments were separated by PFGE for 24 h, stained, and visualized under UV light.

Microarray construction.

The mixed-genome array was constructed using genomic DNA from eight different strains of GAS representing M1 (n = 3; 1 TSS-associated isolate, 1 pharyngitis isolate, 1 isolate from the 1950s), M3 (n = 2; 1 TSS-associated isolate, 1 pharyngitis isolate), M6 (TSS), M12 (TSS), and M28 (TSS). For each M type, equimolar amounts of genomic DNA were mixed and 10 μg of the DNA mixture was ultrasonically sheared (Branson 250/450 sonifier with a 6-mm microtip; output intensity, 1). Fragments of 1 to 1.9 kb were separated on a gel and extracted (QIAGEN). DNA fragments were cloned into pSMART-HC-Kan vectors (Clone-SMART; Lucigen). Ligation products were transformed into electrocompetent Escherichia coli cells (ElectroMAX DH10B; Invitrogen) and plated on kanamycin (30 μg/ml)-containing tryptone yeast plates. A total of 3,840 recombinant clones were arrayed into 96-well plates. Clone inserts were amplified by PCR using SMART primers (Lucigen) with 5′ C₆ amino linkers to facilitate cross-linking to the aldehyde-coated glass slides. Sequencing of amplicons of randomly selected clones confirmed the presence of GAS-specific fragments. Furthermore, probes specific for 34 known virulence genes were obtained by PCR with the primers and templates listed in Table S1 in the supplemental material. PCR products were purified, and the correctness of the size was evaluated on agarose gels for all virulence probes and random DNA inserts. PCR products were dissolved in 3× SSC, pH 7.2 (1× SSC is 0.15 M NaCl plus 0.015 M sodium citrate) and transferred to 384-well flat bottom plates (Nunc) for array printing. Amplicons were arrayed under a controlled atmosphere on CSS-100 silylated aldehyde glass slides (Telechem), with quill pins (Telechem SMP3) in an SDDC-2 Eurogridder (ESI, Canada).

Hybridization.

Genomic DNAs (0.5 μg) of the 76 strains to be tested were labeled with Cy5-dUTP (final concentration, 0.06 mM; Amersham) using the BioPrime DNA labeling system (Invitrogen). Reference DNA (0.5 μg from the mixture used for microarray construction) was labeled with Cy3-dUTP. Arrays were prehybridized (with 1% bovine serum albumin, 5× SSC, and 0.1% sodium dodecyl sulfate) for 45 min at 42°C and washed in MilliQ water. Hybridization was performed overnight at 42°C.

Image analysis and data processing.

For scanning, a ScanArray TM Express (Packard BioScience) was used. Hybridization signals were quantified using ImaGene, version 5.6 (Biodiscovery), software. For all spots, the signal intensity was measured for Cy5 (test strains) and Cy3 (reference), and local background signals were subtracted. Ratio calculations were normalized by correcting for the overall signal intensities in the respective Cy5 and Cy3 channels. Estimated probability of presence (EPP) for genomotyping analysis was done to define cutoff values (30). Briefly, log₂-transformed ratios are fitted to a normal distribution curve to define data sets representing absent and present genes as well as a weighted distribution for markers that cannot be ascribed to one of these two groups with certainty. From a total of 3,874 spotted DNA fragments (3,840 recombinant clones and 34 virulence genes), 2,704 spots (70%, including all virulence genes) yielded a significant Cy3 signal for all strains and were included for further analysis. The remaining 30% of spots induced a rather weak Cy3 signal and did not yield a significant signal for at least one of the strains. A differentiating biomarker was defined as a spot that was absent in at least one of the 76 strains analyzed (i.e., no significant Cy5 signal) after EPP transformation.

Data analysis.

Principal component analysis (PCA) was used as an unsupervised multivariate method to reduce the multidimensional space of data to its principal components (PC) (33). PCA concentrates strongly correlating variables (biomarkers) that vary in similar ways in all experiments into a new variable, a PC. The PC computation, done with MATLAB software (Natick, MA), is displayed as a 2-dimensional (2-D) scatter plot where the position along the axes shows the PCA score of a strain.

Whereas PCA tries to provide a low-dimensional summary of the data, partial least-squares discriminant analysis (PLS-DA) is a supervised multivariate classification method that can be applied to search for a set of biomarkers that distinguish between two defined classes within the total data set. PLS-DA was used to extract a differential gene profile that could distinguish isolates associated with TSS from those that had given rise to superficial manifestations of GAS disease only. Once a PLS-DA model is calculated, it can be validated by prediction of the class membership of samples not used for model construction (17).

Differentiating biomarkers from all strains were hierarchically clustered with The Institute for Genomic Research software (42) (available at http://www.tigr.org/software/tm4) using complete linkage and Pearson's correlation as a distance matrix. PFGE results were imported into Bionumerics software (Applied Maths, Kortrijk, Belgium), and successive hierarchical clustering (complete linkage) was done using Pearson's correlation. Comparison of the genetic resolution of PFGE and the mixed-genome microarray was performed by visual comparison of the dendrograms. Biomarkers that were present or absent in one or two M types (uniquely or commonly present or absent) were subjected to sequencing. To determine the function of a given sequence and integrate it into an appropriate pathway, ERGO bioinformatics was used, as well as BLAST searches in GenBank.

RESULTS

Genetic clustering and differential genome contents of Dutch GAS isolates.

To examine the most pronounced genetic differences among a collection of 76 GAS isolates belonging to M types M1, M3, M12, M28, M4, M6, M11, and M89, this collection was hybridized on a mixed whole-genome microarray. The resulting hybridization patterns were compared using PCA (see Fig. S1 in the supplemental material). From a total of 2,704 biomarkers included for analysis, 366 differentiating biomarkers were found (14%). Of the 366 differentiating biomarkers, 18 were unique to a single M type. The PCA clustered the different strains primarily according to their M types. This indicates that inter-M-type differences outweigh intra-M-type genomic differences. Nevertheless, substantial genetic variation was observed within certain M types, such as M12 and M28. Figure 1 shows the 2-D hierarchical clustering of the 366 differentiating markers with the 76 GAS strains. In accordance with PCA, many of the differentiating biomarkers segregate into separate (M-specific) groups.

FIG. 1.

2-D hierarchical clustering of 76 S. pyogenes strains. The dendrogram on the x axis shows the clustering of the strains. The dendrogram on the y axis shows the clustering of the 366 differentiating biomarkers. Red represents the presence, and green the absence, of a biomarker. Solid bars on the right, genes unique to one or two M types; open bars, genes that are uniquely absent in one or two M types (see Table 2). Twenty-five representative strains, indicated by arrows at the top, were subjected to PFGE (see Fig. S2 in the supplemental material).

Common genetic profiles among TSS-associated GAS strains.

We examined whether the gene content of TSS strains differed from that of non-TSS strains, excluding M-type-related differences. Since TSS strains may have acquired additional (virulence) genes compared to noninvasive strains, both groups were first quantitatively compared for each M type. The average numbers of differential biomarkers and putative virulence factors present in the two disease categories did not differ significantly (Table 1). To examine possible qualitative differences in the gene combinations present in each category, we applied multivariate biostatistics. A total of 52 strains representing 50% TSS and 50% non-TSS strains equally distributed among the different M types were selected at random (the “training set”). These strains were used to construct a PLS-DA model to separate TSS isolates from non-TSS isolates. Although PLS-DA showed a marginal difference in the gene pattern between TSS and non-TSS isolates in the training set, this gene pattern had no predictive value in categorizing the remaining 24 samples (the “test set”) (data not shown). It was therefore concluded that PLS-DA failed to detect a “TSS-specific” gene profile.

TABLE 1.

Presence of differentiating biomarkers per M type in TSS-associated versus non-TSS-associated strains

M type (no. of strains)a	Proportionb of differentiating biomarkersc of the following kind in the indicated category of strains
	Virulence factors (n = 23)		Other biomarkers (n = 343)
	TSS	Non-TSS	TSS	Non-TSS
M1 (15)	47 ± 1.5	42 ± 10.6	68 ± 0.9	64 ± 6.3
M3 (14)	43 ± 0	43 ± 1.8	84 ± 3.1	82 ± 3.7
M12 (14)	49 ± 4.5	47 ± 1.5	66 ± 5.8	62 ± 2
M28 (14)	57 ± 0	56 ± 1.6	54 ± 6.1	52 ± 5
M4 (5)	33 ± 3.1	30 ± 4.3	58 ± 1.2	58 ± 1.8
M6 (5)	41 ± 3.1	46 ± 5	62 ± 1.4	63 ± 1.6
M11 (5)	39 ± 0	39 ± 0	42 ± 0.4	42 ± 0.2
M89 (4)	48 ± 0	46 ± 3.1	43 ± 4.1	42 ± 6.2
Total (76)	46 ± 7.7	46 ± 7.2	64 ± 13.2	61 ± 12.3

^aA total of 38 TSS-associated and 38 non-TSS-associated strains were investigated.

^bExpressed as a percentage ± standard deviation.

^cAll 366 differentiating biomarkers are subdivided into known virulence factors (n = 23) and other differentiating biomarkers (n = 343).

Genetic dissimilarities between TSS and non-TSS M types.

To obtain insight into the genetic differences underlying different M types, M-type-specific biomarker profiles were determined. For this purpose, biomarkers that were uniquely present or absent in one or two M types were identified. Thus, 64 biomarkers were identified (Fig. 1). These biomarkers were sequenced, and their functions were identified using ERGO bioinformatics (Table 2). After function identification, biomarkers representing the same open reading frame were discarded, and the remaining 51 biomarkers were further analyzed. Six of 51 markers (12%) spanned more than one open reading frame. M1 and M3 have traditionally been associated with severe manifestations of GAS disease (12, 14, 24, 39, 43, 52). The presence of M1- and M3-specific biomarker profiles may provide an explanation for this overrepresentation in severe invasive GAS disease. A set of 10 biomarkers (with attributable functions by ERGO) that belonged uniquely to either or both of these M types was identified (Table 2). These biomarkers include three fragments representing phage proteins, the collagen- and fibronectin-binding proteins (cpa and prtf2-like, respectively) (10), a hydrolase of the haloacid dehalogenase superfamily (28), streptococcal inhibitor of complement (sic) (18), the negative regulator of virulence genes (nra) (40), and streptococcal pyrogenic exotoxin A (speA) (26). Finally, a biomarker comprising three gene fragments matching iron(III) binding factor, a hypothetical protein, and O-acetyltransferase was identified.

TABLE 2.

Biomarkers uniquely or commonly present or absent in the M types included in this study^a

Markerb	Presence or absencec in strains of type:								Functiond	Name	M presencee	Contig (nucleotide)	Beginning nucleotide	Ending nucleotide	Strand	Remarksf
	M1	M3	M4	M6	M11	M12	M28	M89
Uniquely present
1	+								Fibronectin-binding protein	prtf2-like	SF370-M1	NC_002737	116923	119064	+
2	+								Iron(III)-binding protein	SPy1063	SF370-M1	NC_002737	872055	873023	+
									Unknown	SPy1064			873123	873467	+
									O-Acetyltransferase (cell wall biosynthesis)	SPy1065			873805	874368	+
3	+								Hydrolase (HAD superfamily)	SPy1066	SF370-M1	NC_002737	874375	875073	+	2*
4	+								Complement inhibitor protein	sic	SF370-M1	NC_002737	1683598	1682675	−
5		+							Phage protein	SpyM3_0723	Phage 315.1	NC_004584	28492	29259	+
									Phage endopeptidase	SpyM3_0724			29259	31307	+
6		+							Phage protein	SpyM3_1451	Phage 315.6	NC_004589	4126	4293	+
									Phage protein	SpyM3_1450			4289	4459	+
									Phage protein	SpyM3_1449			4493	4873	+
7		+							Phage protein	SpyM3_1445	Phage 315.6	NC_004589	5612	6766	+
8		+							Negative transcriptional regulator	nra	M3 SSI-1	NC_004606	111314	109782	−
9		+							Collagen-binding protein	cpa	M3 SSI-1	NC_004606	111745	113976	+
10				+					Phage protein		M6 MGAS10394	NC_006086	1016713	1017393	+
11						+			Unknown	SPs1552	M3 SSI-1	NC_004606	1547974	1547534	−	Match, 118/613
12							+		Phage DNA/RNA helicase	NP_607398.1	M18 MGAS8232	NC_003485	1078848	1077484	−
Commonly present
13	+	+							Pyrogenic exotoxin	speA	M3 SSI-1	NC_004606	583201	582449	−
14	+		+						Phage-associated cell wall hydrolase	SPy1438	SF370-M1	NC_002737	1192514	1191312	−
15	+				+				Unknown	abiR	SF370-M1	NC_002737	49621	51261	+
16	+					+			Phage protein	SpyM3_1302	Phage 315.5	NC_004588	36180	36836	+	Match, 337/954
17	+						+		Pyrogenic exotoxin	speJ	SF370-M1	NC_002737	362130	361435	−
18		+		+					Phage endopeptidase	SpyM3_0724	Phage 315.1	NC_004584	29259	31307	+
19		+		+					Hyaluronoglucosaminidase	hylP.1	Phage 315.1	NC_004584	31307	32311	+	2*
20		+		+					Phage infection protein	SpyM3_0726	Phage 315.1	NC_004584	32324	34225	+	3*
21		+		+					Unknown	SPs0590	M3 SSI-1	NC_004606	619250	618474	−
										SPs0591			619672	619484	−
22			+	+					Transposase		M6 MGAS10394	NC_006086	408121	409116	+
23				+		+			Pyrogenic exotoxin	speH	M6 MGAS10394	NC_006086	744646	745082	+
24							+	+	Fibronectin-binding protein	prtf-15	M6 MGAS10394	NC_006086	159255	161135	+	Match, 1,233/2,801
Uniquely absent
25		−							Transcriptional regulator RofA	rofA	SF370-M1	NC_002737	116531	115041	−
26		−							DNA polymerase IV	dinP	SF370-M1	NC_002737	1532583	1531492	−
27		−							RNA	tRNA-Lys	SF370-M1	NC_002737	1663725	1663797	+
									Acetyltransferase	SPy1994			1663936	1664373	+
28			−						Pyrogenic exotoxin	speG	SF370-M1	NC_002737	190526	191227	+
29			−						Daunorubicin resistance ATP-binding protein DrrA	SPy0518	SF370-M1	NC_002737	419436	420425	+
									Daunorubicin resistance transmembrane protein	SPy0519			420430	421245	+
30			−						Hyaluronan synthase	hasA	M3 SSI-1	NC_004606	1874355	1875539	+
31				−					Phosphohydrolase	SPs1835	M3 SSI-1	NC_004606	1862110	1861796	−
32					−				Short-chain fatty acid transporter	atoE	SF370-M1	NC_002737	126461	127834	+
33					−				Acetate CoA-transferase alpha subunit	atoD.2	SF370-M1	NC_002737	130132	130788	+
34					−				Serine (threonine) dehydratase (tantibiotic biosynthesis)	salB	SF370-M1	NC_002737	1598262	1596640	−
35					−				Ribosomal small-subunit pseudouridine synthase A	SPs0799	M3 SSI-1	NC_004606	788120	788833	+
36						−			Biotin carboxyl carrier protein of oxaloacetate decarboxylase	SPy1176	SF370-M1	NC_002737	966268	966615	+
37						−			Mg2+/citrate complex secondary transporter	SPy1180	SF370-M1	NC_002737	971094	969691	−	2*
38						−			Citrate lyase beta chain	citE	SF370-M1	NC_002737	973897	974781	+	3*
39						−			Citrate lyase alpha chain	citF	SF370-M1	NC_002737	974787	976316	+	3*
40						−			Apo-citrate lyase phosphoribosyl-dephospho-CoA transferase	citX	SF370-M1	NC_002737	976312	976887	+	2*
41						−			Oxaloacetate decarboxylase alpha chain	oadA	SF370-M1	NC_002737	976890	978296	+	2*
42							−		No features known		M3 SSI-1	NC_004606	919614	919730
43								−	Sensor kinase DpiB	SPy1107	SF370-M1	NC_002737	907372	905834	−
									Malate-sodium symport	malP			907533	908861	+
44								−	NAD-dependent malic enzyme	SPy1110	SF370-M1	NC_002737	908895	910058	+
Commonly absent
45	−	−							Fibronectin-binding protein	sfb	M6 MGAS10394	NC_006086	159255	161135	+	Match, 430/690
46	−	−							Fibronectin-binding protein	prtf-1	M6 MGAS10394	NC_006086	159255	161135	+
47				−	−				Hypothetical cytosolic protein	SPy1046	SF370-M1	NC_002737	854757	858860	+
48					−			−	Phage protein	SpyM3_1331	Phage 315.5	NC_004588	13230	13730	+	2*
49					−			−	Phage-encoded transcriptional regulator, ArpU family	SpyM3_1330	Phage 315.5	NC_004588	14184	14615	+	2*
50					−			−	Phage protein	SpyM3_1329	Phage 315.5	NC_004588	15252	15605	+
51						−	−		Hypothetical membrane-associated protein	Spy2174	SF370-M1	NC_002737	1806926	1808002	+

^aUnknown biomarkers were subjected to sequence analysis, and the closest protein match was identified by using ERGO software as well as BLAST searches against GenBank sequences.

^bBoldfaced biomarkers were used for microarray validation by PCR.

^c+, present; −, absent.

^dHAD, haloacid dehalogenase; CoA, coenzyme A.

^eThe best match in ERGO was used (always a >95% match unless otherwise stated in the “Remarks” column). If an identical match was obtained, first phages were identically matched, followed by M types in the order given, beginning with M1.

^f2* and 3* indicate how often biomarkers, corresponding to these genes, were identified. Matches show the number of identical nucleotides/number of nucleotides compared.

Genetic dissimilarities between M types.

Twelve of 51 (24%) markers were phage related, and 18 (35%) represented a (putative) virulence factor. Forty-seven of the biomarkers showed >95% homology to genes present in at least one of the seven completed GAS genome sequences (5, 8, 18, 21, 37, 49, 53). Four markers (markers 11, 16, 24, and 45 in Table 2) showed less than 65% homology to the closest matching gene in ERGO. These four biomarkers include one hypothetical protein, one phage protein, and two putative fibronectin-binding proteins based on protein characteristics as analyzed by ERGO software. M3 and M6 shared three different biomarkers that encompass a 5-kb fragment with >95% homology to phage 315.1 sequences. M11 and M89 strains uniquely lacked three biomarkers covering a 2.4-kb genetic fragment with >95% homology to phage 315.5 sequences. A unique property of all M12 strains was the absence of a functional citrate lyase operon. Based on microarray data, the citE, citF, and citX genes were absent in M12 strains. This finding was confirmed by citE-, citF-, and citX-specific PCRs. The citrate lyase operon may enable bacteria to adapt to metabolic stress conditions, such as low pH, lactate accumulation (35), and nutrient-deprived, hypoxic conditions (50). M12 strains showed growth patterns similar to those of the other M types under different lactate concentrations and acidic conditions (data not shown). However, under nutrient-deprived conditions, the growth of M12 strains was substantially reduced compared to that of other M types (Fig. 2). The mean relative increase in lag time or decrease in maximum growth rate under nutrient-deprived conditions for M12 strains was compared to that for other M types. In comparison with the other M types, M12 strains showed significantly longer lag phases (P < 0.0001) and reduced maximum growth rates (P < 0.05) in diluted medium.

FIG. 2.

Growth comparison between M12 strains (n = 14) and strains belonging to other M types (M1, M3, M28, M4, M6, M11, and M89; three random isolates per M type) under anaerobic, nutrient-rich versus anaerobic, nutrient-deprived conditions. Error bars, standard deviations. All experiments were conducted in duplicate. All OD measurements were normalized by subtracting the initial OD. The lag phase was quantified as the time needed for an increment of 0.2 in the OD. The maximum growth rate was defined as the maximum increase in OD between two measurements (every 20 min). The mean relative increase in lag time or decrease in maximum growth rate for the M12 strains under nutrient-deprived conditions was compared to that for other M types (by an unpaired t test).

Validation of microarray experiments.

The reproducibility of hybridization experiments was controlled using duplicate screening for one random representative of each of the eight M types. All 366 differentiating biomarkers showed 100% reproducible patterns in all the duplicate experiments (data not shown). Twenty-five representative strains (Fig. 1) were subjected to PFGE analysis (see Fig. S2 in the supplemental material). In agreement with the microarray PCA and Pearson correlation clustering, PFGE dendrogram patterns showed M-type-specific clusters and some genetic variation within certain M types.

The presence of 19 virulence factors (see Table S1 in the supplemental material) among the 76 GAS strains was determined by PCR (55). For 5 of these 19 genes, namely, smeZ, cpa, pfbp, prtf-1, and prtf-2, polymorphic sequences are found in the different published GAS genomes. Two different primer combinations per gene (see Table S1 in the supplemental material) were used to PCR amplify these five genes. By using the primers listed in Table S1 in the supplemental material, PCR findings were completely in accordance with microarray data for all the strains and all the virulence factors (data not shown). Likewise, the distribution of 13 randomly chosen markers (10 of which are marked in Table 2) was completely confirmed by gene-specific PCRs for all the strains (data not shown).

DISCUSSION

This study provides insight into the relationship between the GAS genome profile, the M type, and TSS. The strain collection consisted of 76 strains belonging to eight M types that together account for 74% of invasive GAS disease in The Netherlands (57). Microarray data were highly reproducible and completely in line with PCR control experiments. Using this method, gene profiles of GAS strains were examined on three different levels. First, differentiating biomarkers within the whole strain collection were identified. Second, M-type-specific gene profiles were determined. Third, the presence of a common gene profile associated with TSS was explored.

Differentiating biomarkers among the whole strain collection were identified and are visualized in Fig. 1. PCA showed strong clonality for M1, M3, M6, and M11 and a larger degree of genetic diversity for M4, M12, M28, and M89. These findings are in agreement with earlier genotyping studies (11, 15, 16, 31, 36, 47, 51). As expected, genetic variation within the strain collection was largely attributable to phages or phage-like elements (4, 8, 37). The identification of phage elements commonly shared between different M types is suggestive for horizontal transfer of these elements, as has been recently documented (3). The presence of 366 differential biomarkers among a total of 2,704 suggests that roughly 86% of the GAS genome is conserved. This is in agreement with the report of Banks et al., who observed an average genome conservation of 90.6% (4). Dutch M1 and M3 isolates are associated with the severest complication of invasive GAS disease, i.e., TSS (56). We therefore determined which genetic profiles were unique to these M types. Ten different biomarkers were found to be uniquely present in M1 and M3 strains (Table 2). Of these 10 biomarkers, 3 represented phage proteins uniquely present in M3 strains. This underscores the importance of phages and phage elements in the diversification and virulence characteristics of different M types. The seven remaining biomarkers included six virulence factors. The GAS surface proteins Cpa and Prtf2-like are microbial surface components recognizing adhesive matrix molecules (MSCRAMMs), allowing the bacterium to selectively interact with host tissue (10). Iron(III) binding factor is described exclusively for the M1 genome. Mechanisms for iron acquisition are very common among bacterial pathogens and are important for full virulence (58). In a recent publication, an important role for iron binding in the modulation of superantigen expression in GAS was found (27). However, in view of the fact that an O-acetyltransferase gene resides on one biomarker fragment together with iron(III) binding factor, we cannot exclude a hitchhiker effect for this gene. GAS hydrolases are considered to be major virulence factors, playing a role in the release of bacterial surface proteins as well as in the degradation of host tissue (41). sic has been described predominantly for M1 but also for M57 (2). This inhibitor of the membrane attack complex enhances bacterial survival during infection (22). Streptococcal pyrogenic exotoxins (Spe's) have superantigenic properties (1, 26), and the speA gene has been suggested to play a causative role in the development of severe GAS disease, including TSS (44). Overrepresentation of genes in individual invasive M types does not provide proof of their role in the invasiveness of that M type, and the possibility that they are associated only with the particular M type and not with its invasive behavior cannot be ruled out. Whereas the six genes mentioned above are well-known virulence factors (20, 23, 27, 38, 40, 41), it is not clear whether nra and O-acetyltransferase are associated with the pathogenic potential of M1 and M3 strains. nra negatively regulates the expression of cpa (40) and other virulence factors. This factor is not specific for M3, since it has been described for other M types (i.e., M4, M18, M49) not included in this study (40). In addition to M1- and M3-specific virulence factors, we identified four novel biomarkers, including two fibronectin-binding-like proteins, which were unique to one or two M types out of the eight M types included in the study. In addition, this is the first report showing the unique absence of the citrate lyase operon in M12 strains. Citrate lyase is a key enzyme that allows the microorganism to enter the citric acid cycle in the reductive mode. This metabolic “switch” facilitates the survival of the pathogen during environmental transitions encountered in the infective process (50). Preliminary data indeed show that, compared to the other M types in this study, M12 strains have substantially reduced growth under nutritionally deprived conditions. Further research is required to establish whether the other biomarkers identified contribute to specific M type characteristics and thereby to the M type bias in GAS disease.

Finally, a PLS-DA model was used to explore a possible common gene profile among TSS isolates, irrespective of their M type. PLS-DA is the most common method used in comparative genomics and transcriptomics to assess the differences between two different groups within a large data set (17). For all M types, TSS isolates were similar to non-TSS isolates in the total numbers of both differentiating biomarker genes and putative virulence factors present in the strains. In addition, PLS-DA did not yield a specific, predictive set of biomarkers that distinguished TSS from non-TSS isolates. The negative PLS-DA results suggest that there is no common differential gene pattern present among TSS isolates as opposed to non-TSS isolates. This might be relevant to the hotly debated issue about prophylaxis for close contacts of patients with severe invasive GAS disease (48). Other factors, such as differences in expression of virulence genes (7, 54) or host-related factors (32), might be more decisive in the outcome of a GAS infection.

In conclusion, Dutch GAS strains appear to be characterized by unique combinations of commonly shared genes and phage elements. These unique gene combinations contribute to the M type characteristics and possibly the M type bias in GAS disease. An example is the unique absence of the citrate lyase cluster in M12 strains, which may render these strains less fit under nutrient-deprived, hypoxic conditions. In addition, we identified four novel M-type-specific genes, which would not have been identified by “conventional” microarray strategies using previously sequenced fragments only. Furthermore, 10 biomarkers including 6 virulence factors were unique to the M1 and M3 strains in our collection, which may contribute to the extraordinary pathogenic potential of these M types. Finally, we did not find indications for the presence of a common gene profile among strains associated with TSS.