Spread of a newly found trimethoprim resistance gene, dhfrIX, among porcine isolates and human pathogens.

A plasmid-borne gene mediating trimethoprim resistance, dhfrIX, newly found among porcine strains of Escherichia coli, was observed at a frequency of 11% among trimethoprim-resistant veterinary isolates. This rather high frequency of dhfrIX could be due to the extensive use of trimethoprim in veterinary practice in Sweden. After searching several hundred clinical isolates, one human E. coli strain was also found to harbor the dhfrIX gene. Thus, the dhfrIX gene seems to have spread from porcine bacteria to human pathogens. Furthermore, the occurrence of other genes coding for resistant dihydrofolate reductase enzymes (dhfrI, dhfrII, dhfrV, dhfrVII, and dhfrVIII) among the porcine isolates was investigated. In addition, association of dhfr genes with the integraselike open reading frames of transposons Tn7 and Tn21 was studied. In colony hybridization experiments, both dhfrI and dhfrII were found associated with these integrase genes. The most common combination was dhfrI and int-Tn7, indicating a high prevalence of Tn7.

Resistance to trimethoprim, usually plasmid borne, is rather common in clinical human isolates. The resistance is mediated by plasmid genes expressing drug-resistant variations of the target enzyme, dihydrofolate reductase (DHFR) (20). Several different types of genes coding for resistant DHFRs are known. The resistant enzyme variations mediate different levels of resistance (9). The most commonly found gene coding for resistant DHFR is dhfrI borne on Tn7 (1,4,7,19), which was first observed in the R plasmid R483 (1,20). It has recently been established that dhfrI can also be located in an element named integron (23), which can be found in Tn2l-like transposons (28). The integron is recombinationally active and carries several other antibiotic resistance genes, among them dhfrII, dhfrV, and dhfrVII (25,26).
The newly characterized dhfrlX gene (11) has so far been observed only among porcine isolates of Escherichia coli in Sweden, where overall trimethoprim resistance frequency was about 16% in 1991. The rather high frequency of trimethoprim resistance and the appearance of a new resistance gene among porcine E. coli could be regarded as a consequence of the extensive use of trimethoprim (in combination with sulfonamides) in veterinary practice, including in the treatment of pig diarrhea. The dhfrlX gene was originally found in isolates from porcine E. coli collected in 1982 from several farms spread over the southern part of Sweden and sometimes from animals not treated with trimethoprim. The dhfrLX gene was found to be borne on conjugative plasmids mediating resistance to a drug level of about 250 mg/liter. In this study, we wanted to investigate the epidemiology of dhfrIX and other trimethoprim resistance genes among porcine isolates and the possible spread of dhfrlX among human pathogens. To investigate the prevalence of dhfrlX among veterinary isolates, 279 trimethoprim-resistant porcine isolates of E. coli were studied by colony hybridization. Thirty-one of these showed hybridization to the probe for dhfrlX. The ability of the dhfrLX gene to spread could also reflect a risk of its moving into human pathogens. This was investigated in a collection of more than * Corresponding author. 400 human trimethoprim-resistant enterobacterial strains, among which one dhfrIX-positive isolate was actually found.

MATERIALS AND METHODS
Bacterial strains. A collection of 279 trimethoprim-resistant E. coli strains of porcine origin from many farms in different parts of Sweden and isolated at the National Veterinary Institute in the years 1984 through 1989 was studied. All strains varied in serotypes and showed various degrees of trimethoprim resistance, which, however, always corresponded to a MIC of >8 mg/liter. Forty-eight trimethoprim-sensitive strains of porcine E. coli (collected in 1987 through 1989) were used as controls. In a second study, 434 human trimethoprim-resistant enterobacterial strains were studied. A part of these were strains collected in 1989 to 1991 from patients with urinary tract infections (UTI), 97 were from Academic Hospital in Uppsala (Carl Pahlson), 54 were from Danderyd Hospital in Stockholm (Bengt Wretlind), 27 were from Huddinge Hospital in Stockholm, and 44 were from Karolinska Hospital in Stockholm (Signe Ringertz), which also provided 26 trimethoprim-resistant strains collected in Addis Ababa, Ethiopia, in 1986 (17). Forty-six strains were from Finland (Turku and Helsinki) (Elina Heikkila), and 14 were fecal E. coli strains from day care centers in Houston, Tex. (Barbara Murray). Also, 46 Shigella strains were obtained from Bangkok, Thailand (Panida Jayanetra), and 80 other strains of the family Enterobacteriaceae were obtained from Lagos, Nigeria (Adebayo Lamikanra).
DNA probes and labeling procedures. The gene-specific probes are described in Table 1. The dhfrlX gene was represented by the 0.34-kb EcoRV-HindIII fragment, which includes 72 bases upstream of the start codon of dhfrlX (11). As specific probes for the integraselike open reading frames of transposons Tn7 and Tn2l (15,25), the probes int-Tn7 and int-Tn2l were used ( Table 1). The fragment probes were labeled with [a-32P]dCTP by using an Oligolabeling Kit (Pharmacia LKB Biotechnology AB, Uppsala, Sweden). Labeled DNA was purified on Sephadex G-50 columns (Pharmacia LKB). Oligonucleotide probes were 5' labeled with [-y-32P]dATP and T4 polynucleotide kinase (18). Labeled DNA was purified on Sephadex G-25 DNA columns. Colony hybridization. Preparation of filters for colony hybridization was as described earlier (16). Prehybridization and hybridization were done at 42°C in a slowly rotating cylinder. For colony hybridization, strains C600 and JM83 without and with the vector plasmids pBR322 and pUC19 were used as negative controls. Positive controls for the different DNA fragment probes are given in Table 1. For oligonucleotide probes, the positive controls also included pCJO01-1 (Table 1), pLKO221 (26), and pLMO226 (24).
For DNA fragment probe hybridization, filters were washed at 68°C for 30 min, twice in 1 liter of 2x SSC (lx SSC is 0. 15

RESULTS
Occurrence of dhfrIX and other dhfr genes among porcine isolates. DNA probes for different types of dhfr genes, including dhfrlX (cf. Materials and Methods and Table 1), were investigated among the 279 trimethoprim-resistant strains from pigs. As a control, 48 trimethoprim-sensitive porcine E. coli strains were examined with the probe for dhfrIX. No positive hybridizations were observed among these sensitive strains.
The type I dhfr gene was, as expected, most common  a For descriptions of probes used, see Table 1. +, hybridization to probe; -, no hybridization. None of the studied strains hybridized to gene-specific probes for dhfiV, dhfrVII, or dhfrVIII (24).
( Table 2). Of the 279 strains, 166 harbored the dhfrI gene, and in 102 of these (61%), dhfIrwas most likely borne on Tn7 as judged from the concomitant int-Tn7 hybridization. The combination of dhfrI and int-Tn2l hybridizations was observed in 12 strains (4.3%). Twenty-five strains carrying dhfrI hybridized to int-Tn7 and int-Tn2l in combination. This is most likely explained by a Tn7 location of dhfrI and a parallel occurrence of an integron structure carrying int-Tn2l. Six strains hybridized to the probes for dhfrII and int-Tn2l. This group could represent strains in which dhfrII is inserted as a genetic cassette in a Tn21-like structure (25). None of the porcine strains harbored the genes coding for DHFRV, DHFRVII, or DHFRVIII. In 18 strains, hybridizations were observed with the probe for int-Tn7 but not with the gene-specific probe for dhfrI. One group of 24 strains hybridized only to the int-Tn2l probe, and 5 isolates showed hybridization to the probes for both of the integraselike open reading frames of transposons Tn7 and Tn2l but not to any of the dhfr gene probes. Thirty-one of the isolates (11%) showed hybridization to the dhfrlX probe. Among these, one also hybridized to the int-Tn7 probe, and two hybridized to both the dhfrlX probe and the int-Tn2l probe. The dhfrlX gene, which does not resemble a cassette, and the integraselike open reading frames of Tn7 and Tn2l are probably on different locations in these cases. In our previous study, one of the original wild-type isolates, which harbored dhfrlX, was hybridizing to the probes for both dhfrlX and int-Tn2l, but after transfer of dhfrlX into a recipient strain, the transconjugant did not show any hybridization to the int-Tn2l probe (11). The 31 isolates which hybridized to the probe for dhfrlX were collected from 1984 through 1989 in geographically separate areas.
Finally, 29 strains did not hybridize to any of the probes used in this study. These strains thus seem to carry still other trimethoprim resistance genes.
High frequency of trimethoprim resistance is a consequence of extensive use of trimethoprim. Trimethoprim in combination with a sulfonamide was introduced into veterinary practice in Sweden in 1974, and in that year, no trimethoprim-resistant strains could be found among porcine E. coli. As early as 1982, however, 10% of E. coli isolates from pigs with diarrhea were resistant to trimethoprim (5). The rather VOL. 36, 1992 ANTMICROB. AGENTS CHEMOTHER. high frequency of trimethoprim resistance among those isolates is most probably a consequence of the extensive use of trimethoprim-sulfonamides in the treatment of piglets with diarrhea. The use of trimethoprim (in combination with a sulfonamide) mainly for the treatment of UTI in humans has decreased since 1980 (Fig. 1). This decrease was caused by the side effects of the sulfonamide component. The utilization of trimethoprim as a, single drug has not compensated for this decrease (Fig. 1). Furthermore, Fig. 1 shows that the utilization of trimethoprim-sulfonamide increased from 0.56 defined daily doses (DDDs) per 1,000 inhabitants per day in 1973 to 1.35 in 1980 but that it has since decreased to a level of 0.35 DDDs per 1,000 inhabitants per day in 1991. Trimethoprim was used as a single drug for 0.05 DDDs per 1,000 inhabitants per day in 1980, and use increased until 1988. Use has stayed at an almost constant level since then. The utilization of trimethoprim alone amounted to 0.58 DDDs per 1,000 inhabitants per day in 1991. These data are based on sales of trimethoprim and trimethoprim-sulfonamides in DDDs per 1,000 inhabitants per day in Sweden and were obtained from the Swedis Development Center in Uppsala. The use of sulfonamides in single-drug therapy in animals was affected in a similar way ( Table 3). The use of trimethoprim in combination with sulfonamides in veterinary practice has increased every year since 1980 ( Table 3) in spite of the fact that the incidence of neonatal piglet diarrhea has decreased because of the extensive use of highly efficient vaccines against this disease (21). Apparently, this decrease in disease has not been accompanied by the expected decrease in the use of trimethoprim. In consequence, records from the National Veterinary Institute show that the frequency of trimethoprim-resistant strains among porcine E. coli remained at about the same level in 1989 (16%) as in 1982 (10%). All values in Table 3 are from the Swedish Drug Company (Apoteksbolaget), which is a government-owned distributor of all prescription drugs.
Appearance and possible spread of dAfrX among human isolates of trimethoprim-resistant enterobacteria. From a general point of view regarding the use of antibiotics, it is of interest to ask whether the ype IX dhfr gene newly found among porcine isolates will occur and spread among human pathogens. A total of 434 trimethoprim-resistant human isolates were studied by colony hybridization. Of these, 222 were UTI isolates from four Swedish hospitals, while the rest were from Finland, Nigeria, Ethiopia, Texas, and Thailand (see Materials and Methods; Table 4). Among the 434 human isolates, only one gave a clearly positive hybridization signal with the oligonucleotide probe for dhfilX (Table  1). Thus, one human isolate seems to harbor the dhflrX gene (Table 4). This E. coli isolate was from a sample analyzed at a bacteriological laboratory in Uppsala in 1991 and was from a patient with UTI who had no connection to animal farms.

DISCUSSION
Trimethoprim is a clinically useful antibacterial agent because of its selective inhibition of bacterial DHFRs. Structural differences make the human enzyme practically insensitive to the antifolate action of the drug. Bacterial resistance to trimethoprim is mediated by foreign, plasmidborne genes which express drug-resistant variations of DH-FRs. For, more than a dozen of these resistant enzymes, the  (11,13). The origins of the plasmid-encoded enzymes are unknown, but it is reasonable to assume that they were once picked up from environmental organisms naturally resistant to trimethoprim. Some of these resistant enzyme genes have been incorporated into genetic structures like transposon Tn7 (1, 7) and the integron (28), which have effected their efficient spread under the selection pressure of a ubiquitous use of the drug. The emergence of dhfrV could illustrate the plausibility of this argument. This trimethoprim resistance trait seems to have first appeared locally in Sri Lanka, Southeast Asia (29), where it is the dominating type of trimethoprim resistance. Presumably, it has since spread around the world, and a few isolates carrying dhfrV have been found in Europe (25,31). The dhfrV gene is borne on an integron, which could effect its efficient dissemination (25). The dhfrLX gene described here and in a previous paper (11) seems to be an even more clear-cut and more extended illustration of dissemination from a gene reservoir. The dhfrLX gene seems originally to have appeared in strains of E. coli among fecal bacteria in pigs reared in the southern part of Sweden and then spread under the heavy selective pressure of frequent therapeutical use of trimethoprim in veterinary practice. As shown in this paper, it is now widely spread among porcine isolates of E. coli, even those from pig herds with no history of trimethoprim treatment. The dhfLrX gene could have appeared in one animal in one herd and then spread by the extensive trade and transport of piglets between farms in Sweden. This spread is not the dissemination of one single bacterial clone, since the strains carrying dhfrLX belonged to several different serotypes (data not shown). In an earlier study (11), dhfrlX was found on two different conjugative plasmids, which formed efficient vehicles for the spread of resistance under trimethoprim selection pressure. It has been a longstanding debate whether the use of antibacterial agents in animals will lead to resistance genes eventually spreading into human pathogens and to difficulties in the treatment of infectious disease as a sequel. The same kinds of antibiotics are used in both humans and animals, and the environments are not separated. Exchange of drug-resistant enterobacteria between animals and humans is known to occur and to cause disease (8), but it is not well understood how specificity to animal hosts contributes to the biologic containment of, for example, E. coli. Campbell and Mee (2) have shown that the same dhfr gene is carried by identical plasmids in both human bacterial strains and porcine isolates, but it was not possible to determine in which direction the transfer had occurred. In a study performed in a farm environment, where the spread of an E. coli strain with markers could be followed, Marshall et al. (12) showed that an E. coli strain harboring a transferable plasmid rapidly spread among different animal species and to humans even when there was no treatment with antibiotics. Similarly, plasmid-borne streptothricin resistance genes were demonstrated to move from farm animals to humans (10). Streptothricin was used for 2 years for growth promotion in industrial pig farms in a relatively large geographic area of eastern Germany. Streptothricin was not used for any other purpose. Plasmids coding for streptothricin resistance were found in fecal bacteria from pigs which had been fed streptothricin but also in E. coli isolated from humans in direct or indirect contact with the animals and even from people in the community, who had had no contact at all with the farms.
The case of dhfrlX seems to be a similar phenomenon just emerging. The transfer of this trimethoprim resistance trait, which is ubiquitous among porcine isolates, seems to be on the verge of transferring into human bacterial strains. Of the 434 trimethoprim-resistant human isolates investigated, only one, an E. coli strain, showed the presence of dhfrIX. It originated from an elderly patient with UTI living in a city in the middle part of Sweden who had had no farm contacts.
The patient was given several trimethoprim courses of therapy over 10 years.
Trimethoprim has been used extensively in Sweden, mostly for the treatment of UTI. Since all prescription drugs in Sweden are distributed by one state-owned company, it is possible to obtain sales figures representing the total utilization of every drug. Figures for trimethoprim utilization are shown in Fig. 1, where it can be seen that in the years 1980 through 1984, more than 1.25 DDDs were utilized per day per 1,000 inhabitants. This means that statistically, almost 4.5% of the Swedish population was exposed to trimethoprim in those years, provided that the average period of treatment was 10 days. This ought to represent a sizable selection pressure for the spread of trimethoprim resistance.
The distribution of other trimethoprim resistance traits besides dhfrlX in the studied collection of porcine isolates seems to reflect what has been observed earlier among human isolates, among which the most common gene for trimethoprim resistance is dhfrI (7). This gene has mostly been found on transposon Tn7, which seems to have functioned as a very efficient vehicle for the dissemination of this resistance trait. Transposon Tn7 occurs on plasmids like R483 (1) but can also insert itself into the E. coli chromosome. This location seems in fact to be dominating in clinical contexts, since Heikkila et al. (7) have shown that 98% of their dhfrI-carrying E. coli isolates harbored that gene in the chromosome. The dhfrI gene has, furthermore, recently been observed also to occur at a specific insertion site in a recombinationally active genetic structure (28), an integron (23). This integron was earlier shown to contain several other resistance genes (dhfrII, dhfrV, etc.) at the same GTTA locus (25). The 12 dhfrI-carrying strains hybridizing to the probe for int-Tn2l but not to that for int-Tn7 (Table 2) could be examples of this dhfrI location. The 18 strains in Table 2 which hybridized to the probe for int-Tn7 but not to the probe for dhfrI, on the other hand, could represent transposons similar to TnJ825 and Tn1826, which in turn were shown to be very similar to Tn7 in mediating streptothricin (nourseothricin) resistance and in carrying the same VOL. 36,1992 spectinomycin resistance gene (aadAl) as Tn7 but which lack dhfrI (27,30).