Molecular Subtyping to Detect Human Listeriosis Clusters

Brian D. Sauders, Esther D. Fortes, Dale L. Morse, Nellie Dumas, Julia A. Kiehlbauch, Ynte Schukken, Jonathan R. Hibbs, Martin Wiedmann


Emerging Infectious Diseases. 2003;9(6) 

In This Article


From November 1996 through June 2000, a total of 135 L. monocytogenes isolates were collected from human cases, with four mother/newborn pairs of isolates. All four isolates from the newborns matched the subtype of the respective mother and were not included in our analysis. The incidence of reported isolates ranged from 0 to 13 per month, with a median of three. Isolates were seasonally distributed with peaks in July 1997 (n=5), October 1998 (n=13), and September 1999 (n=9). On the basis of the 2000 census population estimate of 10,968,179 [26] for New York State (excluding New York City), we detected a listeriosis rate of 0.33 cases per 100,000. Cases were distributed across the state (Figure 1). Case-patient ages ranged from <1 day to 98 years with a median age of 66 years. Gender was reported for all cases; 76 (58%) of case-patients were female.

A total of 34 ribotypes and 74 PFGE types were differentiated among the 131 human isolates; 19 ribotypes and 50 PFGE types were unique (i.e., represented by only one patient isolate). Ribotypes DUP-1044A and DUP-1052A were each prevalent in >10% of cases, and these two ribotypes alone accounted for 39% of cases. One PFGE type (NY1997ASC0010) accounted for 13% of cases. No other PFGE types accounted for >5% of cases.

Simpson's Index was used to determine the discriminatory power of the subtyping methods used. The D value was 0.923 for ribotyping, 0.975 for PFGE, and 0.980 for combined use of both typing techniques. PFGE further discriminated most ribotypes ( Table 1 ); however, three isolates with indistinguishable PFGE types were further differentiated into two ribotypes ( Table 1 , cluster E).

Ribotyping and PFGE subtyping data were analyzed separately by using a scan statistic on 1- and 3-month windows to detect statistically significant clusters of identical ribotypes and PFGE types. A total of 9 clusters representing 41 (31%) cases were detected by ribotyping, PFGE, or both ( Table 1 and Table 2 ). Clusters were detected throughout the study period (Figure 2). Two clusters (B and G) were epidemiologically linked to national outbreaks and known sources and included 17 (13%) cases. The remaining seven clusters were not epidemiologically defined as outbreaks, and the exact source of exposure was undetermined. Ribotype-based scanning with 1-month windows detected two clusters ( Table 1 , B and E), while scanning with 3-month windows detected six clusters (B,C,E,F,H, and I). PFGE-based scanning with 1-month windows detected five clusters (A,B,E,F,G), while scanning with 3-month windows detected eight clusters (A,B,D,E,F,G,H, and I). All clusters identified by using 1-month windows were also identified by using 3-month windows.

Temporal distribution of listeriosis clusters detected based on ribotype or pulsed-field gel electrophoresis (PFGE) data, using a 3-month window scan statistic. Panels A-G each show the distribution of cases caused by a specific ribotype; ribotypes are denoted in the header of each panel. For panel C, one case caused by ribotype DUP-1044B is included with cases caused by ribotype 116-363-S-2 based on a PFGE match (Table 1, cluster E). Cases, which are part of statistically significant ribotype or PFGE clusters, are denoted by dark bars and labeled by cluster designation (A-I, see Table 1). Open bars indicate cases that were not part of a cluster detected by the scan statistics. Panel H shows human cases, which did not represent clusters and were not caused by any of the ribotypes shown in panels A-G. The X-axis of each panel represents November 1996 to June 2000.

A total of six ribotype-based clusters ( Table 1 ; B, C, E, F, H, and I) of two or more isolates (p≤0.05) were detected, representing a total of 31 (24%) cases. PFGE alone identified eight clusters (A, B, D, E, F, G, H, and I), representing a total of 31 (24%) cases. Ribotyping and PFGE results were used to further refine clusters detected by the scan statistic. All six ribotype clusters contained at least two indistinguishable or closely related (≤3 bands different) PFGE patterns. For the purpose of refining ribotype clusters, we interpreted PFGE patterns differing by ≤3 bands from each other as possibly being clonally related and sharing a recent enough common ancestor to be grouped together for epidemiologic investigations.[27] Three of these clusters (C, E, and I) contained one or more isolates removed from the ribotyped-based cluster because they were considered not closely related to the most common PFGE pattern in the respective cluster (see Figure 3 for two examples of ribotype clusters with multiple PFGE types and two examples of ribotype clusters with indistinguishable or closely related PFGE types). Overall, ribotype clusters that were further supported by indistinguishable or closely related PFGE types represented 26 (20%) cases ( Table 2 ). Of the eight PFGE clusters detected, all, except one (Figure 3, cluster E), comprised isolates with identical ribotypes. Overall, five clusters (B, E, F, H, and I; 23 cases) were detected by the temporal scan statistic on the basis of both ribotype and PFGE data.

Comparison of AscI pulsed-field gel electrophoresis (PFGE) patterns for isolates from selected ribotype clusters. AscI PFGE types are shown for two clusters representing epidemiologically confirmed outbreaks (A and G), one ribotype cluster that was further discriminated by PFGE typing (C), and one cluster with overlapping PFGE and ribotype clusters (E). Isolates with <3 bands difference are shown in bold. The percent similarity does not reflect true phylogenetic distance.

Space-time cluster analysis independently identified three of the ribotype clusters (B, G, and H) and five of the PFGE clusters (B, D, G, H, and I). While some geographic clusters were located within one county (D and G), others comprised cases in one or more counties (B, H, and I). Cluster B comprised two main geographic clusters; one included five cases (Rensselaer and Columbia Counties), and the other included six cases (Broome, Monroe, and Onondaga Counties). An additional three cases from cluster B ( Table 1 ), detected by the temporal scan statistic, were not detected by the space-time analysis (two cases in Albany and one in Erie Counties).