A total of 129 selected isolates of Serratia marcescens which had been
recovered from 50 patients during the 1980-1995 period and which reve
aled phenotypic variation in terms of bacteriocin (phage tail) suscept
ibility, carbon source assimilation, or serotype, were reexamined with
these three phenotypic methods. Seven isolates (5.4%) were bacterioci
n nontypable; all 129 isolates utilized carbon sources and could be se
rotyped. Fourty-eight isolates from 20 patients yielded unambiguous re
sults with these 3 phenotypic methods and were excluded from further a
nalysis. Among the remaining 81 isolates from 30 patients, isolates fr
om 2 patients revealed phenotypic variation in bacteriocin susceptibil
ity only, whereas isolates from 6 patients showed variant bacteriocin
types and variant biochemical profiles, but were of identical serotype
. Isolates from 20 patients revealed variant biochemical profiles only
. Three patients had become superinfected with strains of S. marcescen
s of different phenotype and genotype. In 4 patients, previously motil
e (H12) isolates had become nonmotile (H-). PFGE analysis of XbaI and
SpeI-restricted genomic DNA of the 81 isolates of the 30 patients demo
nstrated the isolates of 22 patients to be genotypically identical. Th
e isolates from 3 patients were closely related by genotype, and those
from an additional patient proved to be possibly related. PFGE analys
is demonstrated one patient to have become infected by two genotypical
ly different strains of S. marcescens of identical serotype, which, ho
wever, differed in bacteriocin type and biochemical profile. It was co
ncluded that PFGE analysis of restricted genomic S. marcescens DNA was
superior to the three phenotypic methods examined comparatively. Sero
typing was more reliable than bacteriocin typing, and the latter techn
ique yielded fewer phenotypic variants than determination of biochemic
al profiles among consecutively recovered isolates from patients with
long-lasting S. marcescens infection.