emm Typing of M Nontypeable Invasive Group A Streptococcal Isolates in Israel

Abstract

We performed emm typing of M nontypeable invasive group A streptococcal (GAS) isolates collected in a prospective population-based study in Israel. One hundred twenty of 131 isolates (92%) had emm sequences compatible with GAS, consisting of 51 different emm types. Eleven isolates were found to be group G streptococcus. Of the 120 isolates, 55 (46%) belonged to 32 types for which there were no typing sera available in the Streptococcal Reference Laboratory in Israel. The other 65 (64%) isolates, consisting of 19 types, had sera available and therefore could have been serotyped. Forty-three isolates had T and emm types which were not correlated according to standard M-typing protocols and were therefore missed. The principal effect of emm typing was the addition of 32 types not previously identified in Israel and the discovery of new associations between emm and T types. emm typing did not significantly change the proportion of M types; the five most common types were 3, 28, 2, 62, and 41. Twenty different types comprised 80% of all isolates. No new emm sequences were discovered. emm typing emphasized the unusually low incidence of M1 strains causing severe disease in Israel. As serological typing of GAS becomes more problematic due to lack of sera and the appearance of new emm types, reference laboratories should replace M typing with emm sequence typing. Development of a GAS vaccine relies on the emm type distributions in different geographical locations. In our study, 7% of isolates (types 41 and 62) are not included in a 26-valent vaccine that is being studied.

Group A streptococcus (GAS) causes a variety of human infections. These range from mild, self-limited diseases like pharyngitis and impetigo to severe, sometimes life-threatening illnesses such as bacteremia, necrotizing fasciitis, and toxic shock syndrome. Typing of GAS has long been the hallmark of both epidemiological studies and the understanding of diseases caused by different strains (2). M protein is a major virulence determinant of GAS that is associated with resistance to phagocytosis, adherence to cells, and virulence in a mouse model of necrotizing fasciitis (1, 24). Serological M typing was developed many years ago and was the only means for typing GAS. Initially, only 50 serotypes were described (12), but later several reference laboratories added some 30 more serotypes (12). This laborious method is becoming obsolete because it is time-consuming and expensive. Many centers have stopped producing specific antisera in rabbits. Sequence analysis of the hypervariable portion of the emm gene encoding M protein (emm typing) has simplified GAS typing and has recently expanded the number of known GAS types from ∼80 to 124 (12).

In various regions of the world, the percentage of M nontypeable strains varies from >90% (20) to <20% (15). The reasons for this variation include technical difficulties and a high prevalence of new emm types, for which serum is not available for M typing (15). In Israel, 67% of 21,517 GAS isolates (mostly from pharyngeal swabs taken over a 10-year period) were M nontypeable (5). In a more recent study, Yagupsky and Giladi found 77% (10 of 13 cases) of GAS strains isolated from children with bacteremia to be M nontypeable (27). In a prospective population-based study of invasive GAS infections in Israel, conducted over a 2-year period, 133 of 409 cases (33%) were M nontypeable. These strains underwent emm sequencing and are the subject of this study.

MATERIALS AND METHODS

Source of GAS strains.

GAS isolates were collected in a population-based study of invasive GAS done in Israel from 1997 to 1999 (19). Isolates were sent to our study center in Jerusalem by 24 of the 25 acute-care hospitals in Israel; 409 isolates were obtained from normally sterile sites and sent for traditional serologic M and T typing to the Israel Ministry of Health Streptococcal Reference Laboratory (MOH), and 402 strains were available for T typing. M typing was performed according to a decision analysis that relies on a known correlation between T and M types and which is similar to that used by Johnson and Kaplan (17). After T typing, only correlated M antisera were used for M typing. T nontypeable strains were not M typed. All strains that were M and T nontypeable were further analyzed by emm typing (3).

emm typing.

PCR of streptococcal isolates was performed according to the recommendations of the Division of Bacterial and Mycotic Diseases, Centers for Disease Control and Prevention (CDC), Streptococcus pyogenes emm sequence database (http://www.cdc.gov/ncidod/biotech/strep/doc.htm). The primers used for amplification of GAS DNA were primer 1 (5′ TATTCGCTTAGAAAATTAA 3′) and primer 2 (5′ GCAAGTTCTTCAGCTTGTTT 3′).

According to the CDC recommendations the sequence of the sense strand of the emm hypervariable coding region was determined. The PCR product was sequenced by automated sequencing, using primer 1 (Hy Laboratory Ltd., Rehovot, Israel). The sequence of bases 65 to 165 was submitted (using the Streptococcal Group A Subtyping Request Form, Blast 2.0 Server) to the National Centers for Disease Control Biotechnology Core Facility Computing Laboratory, where the emm type was determined.

Correlation protocol

T type-M or emm correlation was based upon a preset protocol used by the MOH; this protocol is a combination of the common correlations in Israel and the correlations published by Johnson and Kaplan (17). We also compared T and M or emm correlation to the CDC protocol, which is based on both data from the S. pyogenes emm sequence database (http://www.cdc.gov/ncidod/biotech/strep/emmtypes.htm) and recent publications (2, 4).

Differentiation between GAS and GGS

Differentiation between GAS and group G streptococcus (GGS) was achieved by subjecting S. pyogenes or Streptococcus dysgalactiae subspecies equisimilis to both a pyrrolidonyl aminopeptidase test and a Rapid ID 32 STREP test (bioMerieux, Marcy l'Etoile, France).

Statistical analysis

Statistical analysis was done with SPSS software (release 11.0.1). The χ² test or Fisher exact test was used for differences in proportions where required, and the Mann-Whitney U test was used for nonparametric comparisons. A two-sided P value of <0.05 was considered significant. Two sets of comparisons were made between the isolates. In the first, we compared typeable and nontypeable strains. In the second we compared isolates bearing the same M or emm type to evaluate if there are any parameters attributed to failure of M typing. The comparisons included demographics of patients, the clinical source of the isolate, and the month and year of the study. Comparisons were also made for strains from different regions in Israel by allocating patients to four distinct geographical regions.

RESULTS

emm typing was performed on 131 of the 133 isolates which were nontypeable by M serotyping. The other two strains were not sequenced for technical reasons. All the isolates were PCR amplified, and the resulting sequences matched that of the emm gene by homology analysis. The corresponding emm type was received from the CDC facility. For 11 of 131 emm-typed isolates that were found to have emm genes characteristic of GGS, we repeated the serogrouping. Six of these isolates gave a positive reaction with group G antigen, indicating that the isolates were indeed GGS. The other five gave a positive reaction with the group A antigen. In contrast to this finding, both a pyrrolidonyl aminopeptidase test and a Rapid ID 32 STREP test confirmed that these five isolates were S. dysgalactiae. Thus, the five isolates were GGS presenting a group A antigen.

Among the 120 strains that had emm sequence typing compatible with GAS (Fig. 1), there were 51 different emm types (Table 1). Fifty-five of the 120 isolates belonged to 32 strain types for which there were no typing sera available in the MOH (types 30, 33, 42, 44, 51, 53, 61, 64, 65, 68, 71, 74 to 78, 81, 82, 84 to 87, 89, 92, 94, 95, 103, 113, 118, ST1815, ST3765, and ST5282). This was due to either newly characterized emm types or known types for which antisera were no longer available at the MOH. The remaining 65 isolates, comprising 19 strain types (types 1, 2, 4, 5, 6, 9, 11, 12, 14, 18, 19, 22, 24, 26, 28, 29, 49, 59, and 60), for which sera were available, could have been serotyped by the conventional method. As shown in Fig. 1, for some isolates (49 of 120), the diagnosis should not have been missed, since the T and emm types were correlated. For another set of isolates (43 of 120), the association between the T and emm types did not agree with the protocol (Fig. 1), and therefore the strains could not have been M typed correctly.

FIG. 1.

Number and distribution of GAS isolates in the study. “Traditionally” refers to M or emm types that correlate with the MOH scheme.

TABLE 1.

M and emm types and their corresponding T types

M/emm type	No. of isolates			No. of isolates for which M-emm correlates with T type			New T types associated (no. of isolates)
	Total	M typeda	M NTb−emm typed	Conforms to CDC protocolsc	Conforms to MOH protocolsc	T NT
1	9	5	4	6	6	0	8 (1), 28 (1), 55 (1)
2	28	26	2	23	28	0	6 (5)
3	104	104	0	84	104	0	1 (20)
4	10	4	6	7	8	0	12 (1), 15/17/23 (1), 28 (1)
5	9	6	3	9	9	0
6	5	3	2	2	4	1	2 (2)
9	9	6	3	7	7	0	15/17/23 (1), 28/56 (1)
11	6	3	3	6	6	0
12	11	10	1	10	11	0	11 (1)
13	4	4	0	4	4	0
14	8	2	6	1	4	3	3 (1), 49 (3)
15	5	5	0	0	5	0	15/17/23 (5)
18	9	4	5	4	5	1	5 (1), 6 (1), 12 (1), 15/17/23 (1)
19	5	3	2	0	3	2	8 (1), 15/17/23 (2)
22	7	1	6	5	5	0	15/17/23 (2)
24	1	0	1	0	0	0	15 (1)
25	1	1	0	1	1	0
26	6	0	6	0	0	5	15/17/23 (1)
27	4	4	0	4	4	0
28	49	41	8	45	48	0	4 (3), 49 (1)
29	5	4	1	0	5	0	28/56 (5)
30	1	0	1	0	0	1
33	2	0	2	1	1	1
41	13	13	0	11	13	0	1 (2)
42	1	0	1	0	0	1
44	1	0	1	1	1	0
49	2	1	1	2	2	0
51	1	0	1	1	1	0
53	1	0	1	0	0	0	28/56 (1)
55	1	1	0	1	1	0
59	4	0	4	3	2	1
60	1	0	1	0	0	0	12 (1)
61	2	0	2	2	0	0
62	18	18	0	17	18	0	5/27/44 (1)
64	1	0	1	1	0	0
65	1	0	1	0	0	0	15/17/23 (1)
68	1	0	1	0	0	0	28 (1)
71	1	0	1	0	0	0	14 (1)
74	3	0	3	2	1	1
75	4	0	4	3	3	1
76	4	0	4	3	3	0	4 (1)
77	3	0	3	2	2	1
78	3	0	3	2	2	0	5 (1)
81	4	0	4	1	1	2	49 (1)
82	1	0	1	1	0	0
84	1	0	1	0	0	0	11 (1)
85	2	0	2	0	0	2
86	2	0	2	2	0	0
87	2	0	2	2	0	0
89	1	0	1	1	0	0
92	2	0	2	1	0	1
94	1	0	1	0	0	0	28 (1)
95	1	0	1	0	0	1
103	2	0	2	0	0	2
113	1	0	1	0	0	0	3/13/B3264 (1)
118	2	0	2	1	0	1
ST1815	1	0	1	0	0	0	8 (1)
ST3765	1	0	1	0	0	0	3/13/B3264 (1)
ST5282	1	0	1	0	0	0	3 (1)
Group G	11	0	11	0	0	9	2 (1), 5 (1)
Total	400	269	131	279	318	37	(84)

^aTaken from Moses et al. (19).

^bNT, nontypeable.

^cThe M type or emm sequence correlates with previously known M- and T-type association according to traditional protocols (CDC or MOH).

In the complete survey of 400 isolates, 84 had T-type and emm-M-type correlations different from those usually suggested by the CDC protocols (Table 1). For 41 of the 84 isolates, M typing was performed but the M-T correlations of the MOH were different from those used by the CDC (Table 1). The remaining 43 isolates were designated nontypeable and were therefore emm typed. Among the 120 strains, the five most common emm types, accounting for 24% of the isolates, were types 28, 4, 14, 22, and 26. Interestingly, none of the strains were emm3, indicating that M serotyping succeeded in identifying all M3 isolates. This was also true for a number of other types, including 13, 15, 25, 27, 41, and 62. emm typing of nontypeable isolates did not significantly affect the overall proportions of the different M types: the five most common types before and after emm typing were 3, 28, 2, 62, and 41 (Fig. 2). These accounted for 75.1% of the M typeable isolates (n = 269) but only 51.6% of the isolates typed by emm and M (n = 400). Twenty different types comprised 80% of all 400 isolates.

FIG. 2.

Proportions of the five most prevalent GAS types after emm typing (M and emm typing of 400 isolates).

There was no difference between the M and emm types with regard to the patient's age, the month of isolation, the infected organ, or the hospital or region in Israel. This was also true for strains that should have been typeable according to their T types. Thus, these demographic characteristics could not provide an explanation for the failure to M type these isolates. There was a tendency toward a higher proportion of nontypeable strains in the second year of the study (P < 0.001; Mann-Whitney U test). The nontypeable strains did not correlate with other factors: age of patients, organ infected, hospital of isolation, or region in Israel.

We analyzed the T types of isolates and their correlations to demographical and clinical characteristics of patients. Interestingly, T types 3/13/B3264, 15/17/23, and 28/56 occurred mainly in the first year of the study (P < 0.001), while T type 3 occurred mostly in the second year of the study (P < 0.001). Type 3/13/B3264 was also isolated more frequently in areas outside of the Jerusalem district (26.1 versus 11.3%; P = 0.001). No correlation between T types and other parameters tested was found to be statistically significant.

DISCUSSION

We performed emm typing of M nontypeable isolates collected in a prospective population-based study of invasive GAS infections in Israel. The results of this study demonstrate that a considerable proportion of strains were considered M nontypeable because appropriate antisera were unavailable, either because sera had never been prepared (new serotypes emm82 and above [Table 1]) or because the sera were no longer available (e.g., M53 and M64). Technical problems in the production of high-quality sera or in serotyping may also affect the ability to M type isolates. emm typing of GAS in Israel did not change the distribution of the five most prevalent types. The principal impact of emm typing was the addition of 32 new type strains to the Israel database and the discovery of new associations between emm and T types. There were no newly discovered emm sequences among our strains.

The prevalence of M nontypeable isolates varies among different studies and may be related to the geographical distribution of the types. In a report of 4,760 GAS isolates from Canada (26), 15.4% were nontypeable, while in the United Kingdom (7) only 7.35% of 16,909 strains were nontypeable. In one study in Thailand, 80% were found to be nontypeable (23). This may be due to the presence of newly characterized emm types. Indeed, there are several reports indicating that new emm types appear at a higher rate in certain countries in Asia and Latin America (15, 16). The number of different M and emm types (59 types) found in our cohort was considerably higher than that found elsewhere (10). This is a relatively high number of types, taking into account the small population (6,000,000) and the small area of Israel. This variation may be due to the large number of immigrants arriving in Israel during recent years, originating from widely separated geographical regions throughout the world. Among the four geographical areas in Israel, there was no difference in M or emm type distribution. In one study of invasive and noninvasive disease in Hong Kong, there were a total of 32 M types (13). In a study of throat isolates from children in Rome, there were 22 different emm types (8), while in Aachen, Germany, 216 isolates comprised 18 different emm types (6). In a report from Thailand, 40 invasive GAS strains were found to belong to 24 different emm sequences (22). In a study of throat isolates and isolates from sterile sites in Mexico, there were 31 different emm types (10). Similar to findings in the United States, where among 2,002 invasive GAS isolates the 5 most common types accounted for 49.5% of the isolates (21), in our study the 5 most frequent isolates (Fig. 2) accounted for 51.6% of all strains and 20 types accounted for 80% of the isolates. emm types 3 and 28 belonged to the five most frequent isolates in both countries. In Israel, types 12, 41, and 62 were among the five most common types, whereas in the United States they were not present in the 10 most frequent types (21).

We could not find a correlation between M or emm type and clinical features, such as patient gender, age, or source of infection. However, the power to detect these associations may be limited by the small number of isolates within the specific M types. This is in contrast to the findings of Tyrrell et al., who described clusters of M types as correlating with age (26). We could find no apparent reason for the higher rate of nontypeable strains in the second year of the study. There was no change in laboratory technique, and there was no difference between the frequencies of T nontypeable and T type M or emm discordance in the 2 years of the study.

The unusually low incidence of M1 isolates causing severe GAS disease in Israel (19) was corroborated by emm typing and is considerably lower than that described in many other geographical areas (9, 11, 21, 25).

The association of certain T types with specific M types is a well-known epidemiological observation. This association is the basis for using initial T typing for simplification of the M-typing process (17). We found correlations between T and M types which were previously not considered to associate with each other. This finding, together with unavailable sera at the MOH, explains why 60 of 131 isolates (46%) were M nontypeable. There are several notable emm types (2, 3, 14, 15, 18, 28, and 29) that were found to be associated with unusual T types (Table 1). The association of 20 M3 isolates with T1 is remarkable and is rarely seen in the United States (2, 4, 17). These M3 strains may represent former M1 clones and warrant further study.

Nine of the 11 GGS isolates in our collection were T nontypeable. Five of these 11 GGS isolates were found to have a group A cell wall carbohydrate antigen and group G streptococcal emm sequences, and two of them were T typeable. This phenomenon may be due to transfer of the group A (18) and T antigens from GAS to S. dysgalactiae. The population-based nature of our study may give an indication of the prevalence of this finding in Israel, if we assume that strains that were M typeable would not give a positive M-type result on a GGS. emm sequencing seems to be a reliable means for distinguishing between GAS and GGS.

As serological typing of GAS becomes more problematic due to lack of sera and the appearance of new emm types, reference laboratories should replace M typing with emm sequence typing. For the development of a vaccine for GAS, it is pertinent to know the emm type distribution in different geographical locations. In our study, 7% of strains (types 41 and 62) are not included in a 26-valent vaccine that is being studied (14).