Fitness cost: a bacteriological explanation for the demise of the first international methicillin-resistant Staphylococcus aureus epidemic

Abstract

Objectives

Denmark and several other countries experienced the first epidemic of methicillin-resistant Staphylococcus aureus (MRSA) during the period 1965–75, which was caused by multiresistant isolates of phage complex 83A. In Denmark these MRSA isolates disappeared almost completely, being replaced by other phage types, predominantly only penicillin resistant. We investigated whether isolates of this epidemic were associated with a fitness cost, and we employed a mathematical model to ask whether these fitness costs could have led to the observed reduction in frequency.

Methods

Bacteraemia isolates of S. aureus from Denmark have been stored since 1957. We chose 40 S. aureus isolates belonging to phage complex 83A, clonal complex 8 based on spa type, ranging in time of isolation from 1957 to 1980 and with varyous antibiograms, including both methicillin-resistant and -susceptible isolates. The relative fitness of each isolate was determined in a growth competition assay with a reference isolate.

Results

Significant fitness costs of 2%–15% were determined for the MRSA isolates studied. There was a significant negative correlation between number of antibiotic resistances and relative fitness. Multiple regression analysis found significantly independent negative correlations between fitness and the presence of mecA or streptomycin resistance. Mathematical modelling confirmed that fitness costs of the magnitude carried by these isolates could result in the disappearance of MRSA prevalence during a time span similar to that seen in Denmark.

Conclusions

We propose a significant fitness cost of resistance as the main bacteriological explanation for the disappearance of the multiresistant complex 83A MRSA in Denmark following a reduction in antibiotic usage.

antimicrobial resistance, bacteraemia, competition, spa type, mathematical modelling, regression analysis

Introduction

Methicillin-resistant Staphylococcus aureus (MRSA) have, since the beginning of the 1960s, been implicated increasingly in complicated infections, with higher mortality related to difficulties or delay in initiating effective antibiotic therapy.¹ For more than 30 years Denmark has had a low prevalence of MRSA, with less than 2% MRSA among S. aureus bacteraemias.² However, in the 1960s Denmark, along with several other countries in Europe and Australia, suffered from an MRSA epidemic, and S. aureus bacteraemias caused by MRSA increased rapidly in frequency from <1% of isolates in 1962 to 34% in 1968.³

During the period 1957–73, isolates belonging to the phage group complex 83A dominated as causes of S. aureus bacteraemia infections in Denmark. Complex 83A alone accounted for 24% of all S. aureus bacteraemias in 1969 and 70% of these isolates were multiresistant. The frequency of complex 83A isolates among S. aureus bacteraemias dropped to 6% in 1980, and at this point less than 2% of the S. aureus isolates were multiresistant.⁴ Initially, none of the 83A isolates was methicillin resistant. However, after their initial appearance in 1962, the percentage of complex 83A blood isolates with the MRSA phenotype increased rapidly from 1965 and reached a peak of 76% in 1968, after which it fell, until by 1978 it was below 5% [data from the S. aureus bacteraemia database, Statens Serum Institut (SSI), Denmark]. Many of the phage complex 83A MRSA isolates were also resistant to tetracycline, streptomycin and erythromycin (Figure 1).

Figure 1.

(a) Distribution of antibiotypes of phage complex 83A from blood isolates for the years 1957–80. P, penicillin resistant; PST, penicillin, streptomycin and tetracycline resistant; PSTE, penicillin, streptomycin, tetracycline and erythromycin resistant; PSTEM, penicillin, streptomycin, tetracycline, erythromycin and methicillin resistant; PSTM, penicillin, streptomycin, tetracycline and methicillin resistant; Other, other antibiograms represented by six or fewer isolates per year. (b) Number of methicillin-resistant and methicillin-susceptible S. aureus blood isolates of phage complex 83A in the years 1957–80. Methicillin-R, methicillin-resistant isolates; Methicillin-S, methicillin-susceptible isolates.

The sudden increase in the proportion of MRSA among S. aureus bacteraemias was thought to be due to the spread of a few successful clones of the phage complex 83A.⁵ From 1976 onwards isolates belonging to phage complex 83A were replaced by isolates of phage types 95, group II and the 94,96 complex, which only rarely show resistance to antimicrobials other than penicillin.

A few years after the MRSA epidemic described above, Rosendal et al.⁶ speculated that improved hospital hygiene could have been one factor contributing to the observed reduction. However, the same authors suggested that the changes introduced in antibiotic policy at the time the epidemic peaked could have been a more important determinant, especially if multiresistant strains have a slower growth rate than more susceptible strains.⁶ Thus, if multiresistant MRSA isolates carried a fitness cost that was exposed in the absence of positive selection by antibiotic use, then the reduced use of particular antibiotics could have been a significant factor in the reduced frequency of MRSA isolates following the intervention. However, growth rate was never investigated for the strains isolated from this epidemic and so this hypothesis was never tested.

Today it is known that the acquisition of an antibiotic resistance phenotype often has a negative impact on the fitness of bacteria, resulting in reduced growth rates.^7,8 The staphylococcal cassette chromosome mec (SCCmec) has been reported to cause high levels of oxacillin resistance in S. aureus at the expense of a reduced growth rate.⁷ Furthermore, others have reported reduced fitness of bacteria caused by various antibiotic resistance types.^9,10 In this study we asked whether historic S. aureus isolates from the Danish epidemic with the MRSA phenotype showed evidence of reduced fitness associated with being multiresistant. If this were the case, it might explain the disappearance of MRSA in Denmark during the late 1970s, when the selection pressure caused by usage of particular antibiotics was reduced.⁶

The isolates in this study were chosen based on phage typing and spa typing. Phage typing was the method originally used for typing S. aureus, and all isolates in the study were of phage complex 83A, the dominant phage type of the period dealt with in this study. The present typing method for S. aureus isolates is spa typing and spa types can be grouped into different clonal complexes (CCs). Isolates in this study all had a spa type related to CC8.

Materials and methods

Selection of bacterial isolates

Denmark has a unique collection of S. aureus isolates at SSI from nearly all patients with bacteraemia, including clinical data, which have been systematically collected and stored since 1957, and the collection now comprises >40 000 strains. Note that approximately half of Danish S. aureus bacteraemias are community acquired. Forty isolates of phage complex 83A covering the period 1957–80 were selected for this study, based on their spa type [all inferred to belong to the multilocus sequence typing (MLST) group CC8] and to represent the different resistance phenotypes isolated during this period (Table 1). The isolates were freeze-dried upon collection and were stored under these conditions until 2007–08, when they were grown, frozen and re-stocked in vials containing 10% glycerol at −80°C. Thus, the strains have been stored under conditions where growth cannot occur, making it highly unlikely that any genetic alterations have occurred since their original isolation. Nevertheless, we used contemporary susceptible strains for the comparative fitness studies in order to reduce the possibility of bias due to loss of fitness by prolonged storage under freezing conditions. spa typing was performed as previously described.¹¹ The relative fitness of each of the 40 isolates was determined in growth competition assays using a clinical bacteraemia isolate, E11147, as a reference isolate. E11147 is a penicillin-resistant S. aureus isolated in 1980; it belongs to phage group 80 and has a growth rate similar to penicillin-resistant 83A isolates [determined by measuring the change in optical density (OD) over 16 h, reading OD₄₉₅ every 5 min]. E11147 could be easily distinguished from the isolates of 83A by the pigmentation of its colonies (forming greyish colonies compared with the yellow colonies of complex 83A isolates). In addition, 10 of the investigated 83A isolates were also tested in growth competition assays with a different reference isolate, E3951 (phage type 83A), to ensure that the fitness tendencies observed with E11147 were independent of the particular reference isolate. E3951 could be distinguished because it was susceptible to the tested antibiotics.

Table 1.

Details of isolates in this study and results of competition assays

Isolate	Isolation year	Antibiograma	spa type	Relative fitness, fitt ± standard error	P valueb	nc
E3008	1967	susceptible	t008	0.99 ± 0.009	0.4	6
E3951	1969	susceptible	t008	1.01 ± 0.005	0.03	6
E5276	1971	susceptible	t008	1.00 ± 0.012	0.7	6
E5755	1972	susceptible	t008	0.99 ± 0.014	0.6	6
E11097	1980	susceptible	t008	1.02 ± 0.003	0.002	6
E267	1957	P	t008	0.99 ± 0.010	0.3	6
E676	1959	P	t008	0.99 ± 0.013	0.5	6
E1607	1963	P	t008	1.03 ± 0.011	0.03	12
E2129	1964	P	t008	0.99 ± 0.014	0.5	6
E4397	1969	P	t008	0.99 ± 0.013	0.8	6
E7497	1974	P	t008	1.02 ± 0.009	0.02	12
E10605	1980	P	t008	1.03 ± 0.011	0.01	12
E228	1957	PS	t008	0.96 ± 0.009	0.001	12
E4886	1970	PS	t008	0.94 ± 0.013	0.007	6
E213	1957	PST	t008	0.99 ± 0.007	0.3	10
E1210	1961	PST	t008	0.98 ± 0.014	0.2	9
E1421	1962	PST	t008	0.93 ± 0.008	<0.0001	9
E3069	1967	PST	t211	0.98 ± 0.008	0.08	12
E3361	1967	PST	t008	0.98 ± 0.010	0.1	12
E3638	1968	PST	t008	0.96 ± 0.014	0.008	15
E6479	1973	PST	t051	0.96 ± 0.008	0.003	9
E10001	1979	PST	t051	0.97 ± 0.006	0.003	6
E1272	1961	PSTE	t024	1.00 ± 0.008	0.9	10
E3994	1969	PSTE	t051	0.98 ± 0.009	0.03	6
E2125	1964	PSTM	t051	0.95 ± 0.010	0.005	6
E2522	1965	PSTM	t051	0.97 ± 0.007	0.003	12
E2600	1966	PSTM	t008	0.92 ± 0.018	0.007	6
E3275	1967	PSTM	t051	0.97 ± 0.018	0.2	12
E3607	1968	PSTM	t051	0.98 ± 0.013	0.09	12
E3647	1968	PSTM	t051	0.95 ± 0.011	0.007	6
E7192	1974	PSTM	t051	0.96 ± 0.014	0.03	12
E9001	1977	PSTM	t051	0.96 ± 0.007	0.002	6
E9770	1978	PSTM	t054	0.93 ± 0.005	<0.0001	6
E10037	1979	PSTM	t051	0.98 ± 0.008	0.05	12
E2214	1965	PSTEM	t051	0.96 ± 0.011	0.004	10
E2682	1966	PSTEM	t008	0.93 ± 0.009	0.0009	6
E4480	1970	PSTEM	t211	0.94 ± 0.012	0.003	6
E4879	1970	PSTEM	t2482	0.95 ± 0.010	0.004	6
E6562	1973	PSTEM	t051	0.86 ± 0.017	0.001	5
E7132	1974	PSTEM	t008	0.90 ± 0.011	0.0002	6

^aP, penicillin; S, streptomycin; T, tetracycline; E, erythromycin; M, methicillin.

^bP value representing difference of average fit_t from 1 (the reference isolate).

^cn, number of fit_t, which the average fit_t was calculated from.

Susceptibility testing

Susceptibility to penicillin, streptomycin, tetracycline, erythromycin and cefoxitin was determined by tablet diffusion with Neo-Sensitabs (Rosco Diagnostica, Taastrup Denmark) on Danish Blood Agar (SSI Diagnostica, Hillerød, Denmark). Interpretation of the results was based on guidelines by the manufacturer.¹²

MICs were determined for isolates proven streptomycin non-susceptible by tablet diffusion. This was performed using Etest^® (bioMérieux, Herlev, Denmark), following the recommendations of the manufacturer.

Genotype analysis

The presence of mecA was determined by PCR as previously described.¹¹ The presence of erythromycin and tetracycline resistance genes was determined using a multiplex PCR as described by Strommenger et al.¹³ with the following modifications: a 50 μL reaction volume contained 0.2 μM of each primer, 0.1 μM dNTP, 1.5 mM MgCl₂, 2.5 U of Amplitaq DNA polymerase (Invitrogen) in 1× PCR buffer (Invitrogen) and 5 μL of template DNA.

Streptomycin resistance genes ant(6)-Ia and str were investigated by multiplex PCR as described by Hauschild et al.¹⁴ with the following modifications: a 50 μL reaction volume contained 0.1 μM dNTP, 1.5 mM MgCl₂, 2.5 U of Amplitaq DNA polymerase in 1× PCR buffer (Invitrogen), 0.2 μM of each primer and 5 μL of template DNA.

SCCmec typing was performed as described by Kondo et al.¹⁵ where applicable and as described by Oliveira and de Lencastre¹⁶ when typing by the method of Kondo et al.¹⁵ was not applicable.

Competition assays

Competitive growth of the reference isolate and the 40 S. aureus complex 83A isolates was performed as described by Sander et al.¹⁷ with minor modifications. Initially, 4 mL of Luria broth (LB) was inoculated with the reference isolate and one of the 40 test isolates, each at a final concentration of 10³ cfu/mL. After 24 h of incubation at 35°C with 150 rpm shaking (1 cycle of competition, approximately 17 generations), dilution and plating were performed on 5% blood plates (SSI Diagnostica, Hillerød, Denmark) without antibiotics. Hence, distinction of the reference isolate and clinical isolates was performed based on colour differences of the colonies; the reference isolate formed white-greyish colonies, as opposed to the yellowish pigmentation of the 83A isolates. After 24 h of incubation the plates were placed in a refrigerator for 24 h in order to enhance the colour difference between the reference isolate and the test isolate. Successive cycles of growth competition were performed by transferring cells from a competition tube to fresh LB at a final concentration of 10³ cfu/mL. Each competition assay was performed in duplicates of up to four competition cycles and serial dilutions were plated in duplicate.

Calculation of the difference in fitness was performed using the following function as described by Sander et al.:¹⁷ 

where r_t and s_t denote the number of drug-resistant and drug-susceptible cells, respectively, at a given time t and r_t−1 and s_t−1 denote the number of drug-resistant and drug-susceptible cells, respectively, at the preceding timepoint. The quotient of the ratios of the cell numbers was standardized with the exponent 1/17 because cell numbers were determined approximately every 17th generation. The data are presented as relative bacterial fitness (fit_t), defined by Sander et al.¹⁷ as fit_t= 1 + S_t. A fit_t of 1 represents identical competitive fitness to the reference isolate, whereas fit_t< 1 indicates decreased competitive fitness compared with the reference isolate.

Additional experiments with E3951 (phage type 83A) as the reference isolate were performed as described above, with the modification that plating was performed on 5% blood plates with and without 0.5 mg/L penicillin, to distinguish between E3951 and the test isolates.

Mathematical modelling

We applied a mathematical model of hospital and community transmission of S. aureus originally developed by Austin et al.^18,19 The model takes into account the transient colonization of hosts as well as antibiotic treatment-induced resistance, with the latter term set to zero because the resistance is plasmid-borne. The basic reproductive numbers of susceptible (S) and resistant (R) strains, R₀ and R₀′, respectively, are related to the contact or transmission rates (β, β′) and the durations of colonization (μ) and treatment (f).¹⁹ The parameter values for rates of transmission (β), duration of colonization (μ) and duration of treatment (f) used in the model are based on Austin et al.¹⁹ This cost of resistance is expressed mathematically as the transmission fitness R₀′/R₀, where R₀′ and R₀ represent the reproductive numbers of the resistant and susceptible strains and where R₀ = β/μ + f and R₀′ = β′/μ′ + f′. A flow diagram describing the model is shown in Figure S1 (available as Supplementary data at JAC Online) and the parameters are defined in Table S1 (available as Supplementary data at JAC Online). The reader is referred to Austin et al.¹⁹ for a detailed description of the model.

Statistical methods

The t-test (P < 0.05) for the slope parameters was applied to differences in competitive fitness of isolates in correlation to year of isolation, within groups of isolates with identical antibiograms. Additionally, t-tests (P < 0.05) for the slope parameters were performed to test for a correlation between the number of antibiotic resistances and the relative competitive fitness of the isolates. To test the significance of differences in fitness level from 1, one-sample t-tests were performed for each isolate. Two-tailed t-tests and Mann–Whitney U-tests were performed, where applicable, to test whether fitness results differed significantly depending on which of the two reference isolates (E11147 or E3951) was used. All the above tests were performed using GraphPad Instat 3 (GraphPad Software Inc., San Diego, CA, USA).

Multiple regression analysis was performed using Statistica 5.5 (Statsoft Inc., Tulsa, OK, USA) with the difference in competitive fitness as the dependent parameter and resistances to penicillin, streptomycin, tetracycline, erythromycin and methicillin as independent parameters.

Results and discussion

The individual antibiotic resistance phenotypes of the 40 isolates belonging to complex 83A, the majority of which had spa type t008 or t051, showed that some isolates were susceptible to all tested antibiotics, whereas others carried resistance towards one, two, three, four or five different antibiotics (Table 1). Tetracycline-resistant isolates carried both tet(M) and tet(K), apart from one isolate, E1421, which only carried tet(K). Erythromycin-resistant isolates carried erm(A) and all were negative for erm(C). Streptomycin-resistant isolates all had very high MIC values (>1024 mg/L, with only two exceptions, 128 and 24 mg/L), suggesting that the phenotype was caused by ribosomal mutations. In agreement with this, all streptomycin-resistant isolates were negative for ant(6)-Ia and str. Methicillin-resistant isolates all carried SCCmec type 1.

In growth competition assays, the majority of the isolates showed lower fitness than the reference isolate (Figure 2). Exceptions, which showed significantly higher fitness values than the reference isolate, E11147, were two of the five susceptible isolates and three of the seven isolates that were only resistant to penicillin (Table 1). Competition assays were also performed for 10 of the isolates using a phage type 83A isolate, E3951, as reference. There was no significant difference in relative fitness in assays performed with the E11147 or E3951 reference isolate (Supplementary Data, available as Supplementary data at JAC Online). This indicates that the choice of reference isolate did not influence the outcome of the competition experiments.

Figure 2.

Relative fitness in relation to year of isolation. Isolates of phage complex 83A in the period 1957–80 and fitness compared with the reference isolate. P, penicillin resistant; PS, penicillin and streptomycin resistant; PST, penicillin, streptomycin and tetracycline resistant; PSTE, penicillin, streptomycin, tetracycline and erythromycin resistant; PSTM, penicillin, streptomycin, tetracycline and methicillin resistant; PSTEM, penicillin, streptomycin, tetracycline, erythromycin and methicillin resistant.

The reduced fitness of the multiresistant isolates was statistically significant in the great majority of cases (Table 1) and the magnitude of this reduction was independent of the year of isolation (Figure 2 and Table 1).

There was a significant negative correlation between the number of antibiotics to which an isolate was resistant and its fitness relative to both the reference isolate and to the fully susceptible isolates (P < 0.0001) (Figure 3 and Table 1). The uniformity in the nature of the specific resistance mechanisms carried by these isolates for each individual antibiotic argues against the presence of different types of resistance mechanisms as a major cause of differences in relative fitness. We concluded from the correlation (Figure 3 and Table 1) that the accumulation of genetic alterations leading to a multiple antibiotic-resistance phenotype imposed on average an increased fitness cost on S. aureus bacteraemia isolates. There was only one exception to this correlation, namely that isolates resistant to penicillin, streptomycin and tetracycline on average had a lower fitness than isolates resistant to penicillin, streptomycin, tetracycline and erythromycin, despite being resistant to one antibiotic less (Figure 3 and Table 2). We also note that there is no evidence from the data that the relative magnitude of the fitness costs became ameliorated by compensatory evolution over the period covered by this study.

Table 2.

Relative fitness, fit_t, and standard error grouped on the basis of resistance phenotypes

Phenotype	Mean fitt	Standard error
Susceptible	1.00	0.005
P	1.01	0.008
PS	0.97	0.010
PST	0.97	0.007
PSTE	0.99	0.012
PSTM	0.96	0.006
PSTEM	0.92	0.016

P, penicillin resistant; PS, penicillin and streptomycin resistant; PST, penicillin, streptomycin and tetracycline resistant; PSTE, penicillin, streptomycin, tetracycline and erythromycin resistant; PSTM, penicillin, streptomycin, tetracycline and methicillin resistant; PSTEM, penicillin, streptomycin, tetracycline, erythromycin and methicillin resistant.

Figure 3.

Distribution of phage complex 83A isolates according to number of resistances and fitness compared with the reference isolate. P, penicillin resistant; PS, penicillin and streptomycin resistant; PST, penicillin, streptomycin and tetracycline resistant; PSTE, penicillin, streptomycin, tetracycline and erythromycin resistant; PSTM, penicillin, streptomycin, tetracycline and methicillin resistant; PSTEM, penicillin, streptomycin, tetracycline, erythromycin and methicillin resistant. There was a negative correlation between number of resistances and difference in fitness (P < 0.0001, R²= 0.45).

Multiple regression analyses showed that streptomycin resistance (P = 0.05) and methicillin resistance (P = 0.03) were each significantly correlated with reduced fitness in the isolates. These correlations are in agreement with other studies showing that methicillin resistance⁷ and streptomycin resistance²⁰ are associated with reduced bacterial fitness.

We propose that the reduced fitness of S. aureus isolates, shown here to be quantitatively correlated with multiple antibiotic resistances, is a possible explanation for the reduction in MRSA during the 1970s. The initiating factor for this change could have been the change in policy introduced in Denmark in 1969, which included improved infection control measures and reduced antibiotic usage, in particular of tetracycline and streptomycin.⁶ This change in policy was associated with a reduced prevalence of streptomycin-, tetracycline- and methicillin-resistant isolates. Paradoxically, given that the frequency of MRSA declined, the antibiotic usage policy did not include a reduction in methicillin consumption itself. Furthermore, there is no evidence that methicillin or any of the isoxazolyl penicillins select for MRSA, while the antibiotics often selecting for MRSA are the fluoroquinolones, macrolides and cephalosporins,^21,22 and, in the 1960s and 1970s, tetracycline and streptomycin.⁶ Thus, we suggest that the reduction in the frequency of MRSA was due to (i) a reduction in the selection pressure by tetracycline and streptomycin and (ii) a fitness cost, measured in this study, associated with the presence of multiple antibiotic resistances.

We applied a mathematical model of hospital and community transmission of S. aureus strains, originally developed by Austin et al.,^18,19 to test the prediction that a combination of the effects of biological fitness costs associated with resistance and a reduced selective pressure associated with the reduced use of specific antibiotics could account for the observed reduction in the frequency of MRSA within the relevant period after the intervention. The initial frequency of resistance shown in Figure 4, 0.225, is based on the maximum observed frequency of complex 83A MRSA S. aureus bacteraemias in 1967–68. Simulations of the frequency of antibiotic resistance over the following period predict that biological fitness costs in the range observed (from 1% to 7% per cell generation) are large enough to result in reduced frequencies of resistance, and that the magnitude of the reduction would be susceptible to the level of antibiotic selective pressure (compare Figure 4a with 4b, where antibiotic consumption is set to differ by a factor of 2, from 6% to 3% of the population). Note also that in the absence of a fitness cost, and in particular at the higher level of antibiotic consumption, the frequency of MRSA bacteraemias would be predicted to remain at a high level (Figure 4). Thus, the simulations suggest that biological fitness costs of the magnitude measured for most of the MRSA isolates in 1967–68 would have been sufficiently large to reduce the frequency of MRSA to approximately the levels observed in Denmark, and within the observed time frame of approximately one decade. As a caveat, the model uses several assumed parameters, albeit approximated from the dataset used by Austin et al.¹⁹ The agreement between the model simulations and the epidemiological data does not rule out the possibility that other factors, including improved hospital hygiene and other interventions, may also have played a significant role. Indeed, a very interesting study of MRSA in 38 French hospitals has shown that a sustained programme of patient isolation and hand hygiene, coupled with both passive and active surveillance, was associated with a significant and progressive decrease in MRSA frequency (−35%) from 1993 to 2007.²³

Figure 4.

Frequency of hosts infected or colonized with MRSA (y-axis) as a function of time in months. The initial frequency (set at 0.225) is an approximation of the observed frequency of 83A MRSA bacteraemias in 1968 in Denmark at the peak of the epidemic. (a) Predicted changes in frequency as a function of biological fitness cost with antibiotic consumption set at 6% of the population. (b) Predicted changes in frequency as a function of biological fitness costs with antibiotic consumption set at 3% of the population.

Isolates from the Danish epidemic in this study have been shown to carry a biological fitness cost, which, with the lowered consumption of tetracycline and streptomycin, could explain their disappearance. Isolates with the following three resistance antibiograms were the most common during the epidemic (Figure 1): penicillin, streptomycin and tetracycline; penicillin, streptomycin, tetracycline and methicillin; and penicillin, streptomycin, tetracycline, erythromycin and methicillin. In this study these isolates were shown to carry the largest fitness cost. This supports the suggestion that the reduction in antibiotic selection pressure due to the change of policy from 1969 exposed the biological fitness costs of these multiresistant MRSA strains and contributed to a decrease in their frequency.

The epidemic of MRSA in 1965–75 was an international epidemic. The increase and subsequent decrease in prevalence of MRSA isolates in this period was not only observed in Denmark; several countries, including France, England and Australia, reported the same trend,^1,24–28 although in these other countries the frequency of MRSA subsequently increased.^27,29 Unfortunately there is only modest information from these epidemics, in particular regarding antibiotic consumption. The isolates from this international epidemic have in many cases been characterized with respect to phage types, but in most cases they have not been further characterized in detail, as the Danish isolates have been, e.g. with respect to growth rate and resistance genotypes. One study from England in 1974 reported multiresistant S. aureus to belong to phage group I or III or complex 83A in the 1960s and many of these were multiresistant.³⁰ In the hospitals where tetracycline consumption was reduced¹ there was a parallel decrease in the number of multiresistant isolates. This decrease occurred in spite of constant or increasing use of methicillin or cloxacillin.¹ It is plausible that the English MRSA isolates of this time also carried a fitness cost, as the reduction in the number of MRSAs was associated with the change in antibiotic consumption.

A recent prospective study from Sweden³¹ found only a very slight reduction in resistance to trimethoprim among Escherichia coli isolates in spite of a major decrease in trimethoprim consumption over the 2year period of the study. The authors suggested that a possible reason for the poor response was that the trimethoprim-resistant E. coli did not suffer any significant fitness cost when compared with trimethoprim-susceptible isolates. In contrast, a high fitness cost for resistant mutants has been suggested as an explanation for the low frequency of E. coli resistant to fosfomycin in Italy and France.³² These correlations suggest that it may be useful to make prospective determinations of fitness cost in resistant outbreak bacteria and employ the information to help determine which type of intervention could be most effectively employed to prevent further increases in resistance.

In conclusion, we suggest that the fitness costs associated with multiple antibiotic resistances could have been an important factor contributing to the disappearance of multiresistant MRSA of phage complex 83A during the epidemic in 1965–75.

Funding

This study was conducted as part of the EU-FP6 project, EAR (LSHM-CT-2005-518152) and the EU-FP7 project, PAR (HEALTH-F3-2010-241476). K. I. U. acknowledges funding from Carl Bennet AB, and D. H. acknowledges support from Vetenskapsrådet, SSF, Vinnova and the SSAC. The funders had no role in study design, data collection and analysis, decision to publish or preparation of the manuscript.

Transparency declarations

None to declare.

Acknowledgements

Frank Hansen, Lone Ryste Hansen Kildevang and Julie Hindsberg Nielsen are acknowledged for excellent technical assistance. Dr T. Hauschild is thanked for providing positive controls for the streptomycin resistance PCRs.