Antimicrobial resistance is not new, but the number of resistant organisms, the geographic locations affected by drug resistance, and the breadth of resistance in single organisms are unprecedented and mounting1. Diseases and disease agents that were once thought to be controlled by antibiotics are returning in new leagues resistant to these therapies. In this review, we focus on the underlying principles and ecological factors that affect drug resistance in bacteria. It should be stressed, however, that antimicrobial resistance is also evident in other microorganisms—namely, parasites, fungi and viruses2.

Drug-resistant strains initially appeared in hospitals, where most antibiotics were being used3. Sulfonamide-resistant Streptoccoccus pyogenes emerged in military hospitals in the 1930s4. Penicillin-resistant Staphylococcus aureus confronted London civilian hospitals very soon after the introduction of penicillin in the 1940s5. Similarly, Mycobacterium tuberculosis with resistance to streptomycin emerged in the community soon after the discovery of this antibiotic6.

Resistance to multiple drugs was first detected among enteric bacteria—namely, Escherichia coli, Shigella and Salmonella—in the late 1950s to early 1960s7,8,9. Such strains posed severe clinical problems and cost lives, particularly in developing countries. Nevertheless, the resistance problem was perceived by some, most notably those in the industrialized world, as a curiosity of little health concern confined to gastrointestinal organisms in distant countries. This attitude changed in the 1970s when Haemophilus influenzae and Neisseria gonorrhoeae, organisms that cause respiratory and genitourinary disease, respectively, emerged with resistance to ampicillin10,11 and, in the case of Haemophilus, with resistance to chloramphenicol and tetracycline as well12,13. Fueled by increasing antimicrobial use, the frequency of resistance escalated in many different bacteria, especially in developing countries where antimicrobials were readily available without prescription. Poor sanitation conditions aided spread and small healthcare budgets prevented access to new effective but more expensive antibiotics1. Since the 1980s, a re-emergence of tuberculosis has occurred that is often multidrug resistant (MDR) and enhanced by human immunodeficiency virus infection14. The severity of and difficulty in treating MDR strains necessitates the use of several, sometimes six to seven different, drugs15.

Key problems of resistance in hospitals and communities

Multiply resistant organisms render therapy more precarious and costly (Box 1) and sometimes unsuccessful. Individuals may succumb to MDR infections because all available drugs have failed, especially in the developing world1. Notable global examples include hospital and community MDR strains of Mycobacterium tuberculosis, Enterococcus faecium, Enterobacter cloacae, Klebsiella pneumoniae, S. aureus, Acinetobacter baumanii and Pseudomonas aeruginosa3,16,17,18 (Box 2, World Health Organization website). In developing countries, MDR enteric disease agents such as Salmonella enteritidis, Shigella flexneri and Vibrio cholerae threaten and circumvent public health measures.

Overall, in the United States and the United Kingdom, 40–60% of nosocomial S. aureus strains are methicillin-resistant (MRSA) and usually MDR17,18. More deaths are associated with MRSA than with methicillin-sensitive strains19. A steadily increasing, small proportion of MRSA also now shows low-level resistance to vancomycin (the drug of choice), leading to treatment failure20,21. And in three US states, full-fledged vancomycin-resistant strains of S. aureus have appeared22,23, having acquired the resistance trait from vancomycin-resistant enterococci. The latter, in particular MDR strains of E. faecium, have plagued clinicians treating immunocompromised individuals in hospitals in the United States and elsewhere for more than a decade24,25. At present, the newly developed drugs daptomycin, linezolid and the streptogramin combination, dalfopristin/quinopristin, can treat vancomycin-resistant enterococci, MRSA and vancomycin-resistant S. aureus, although some strains have emerged with resistance to the latter two agents26,27.

Among the Gram-negative bacteria, hospital infections caused by P. aeruginosa and A. baumanii are sometimes resistant to all, or all but one, antibiotics, which seriously challenges the treatment of immunocompromised individuals and can result in death3. The extended-spectrum β-lactamases, carried among Enterobacteriaceae such as Enterobacter and Klebsiella, destroy even the latest generations of penicillin and cephalosporins28,29,30. Of particular note is the increase in strains bearing metallo-β-lactamases that inactivate carbapenems—drugs that are often the 'last resort' in serious infections of Gram-negative bacteria31,32.

The community has become similarly encumbered with MDR organisms. Some strains of E. coli, a common cause of urinary tract infection, resist members of six drug families including the more recently recommended fluoroquinolones. In parts of southeast Asia and China, 60–70% of E. coli are resistant to fluoroquinolones33; in the United States and other industrialized countries, frequencies approaching 10% are also worrisome because the trend jeopardizes the value of this drug family34,35.

Resistance in pneumococci continues to be an ever-increasing global threat that curtails treatment of pneumonias and ear infections, particularly in children. Having started with penicillin resistance, the organisms now tout resistance to macrolides and tetracyclines in many areas36. One study has predicted that multidrug resistance—will override single-drug resistance in the present decade37. Strains of N. gonorrhoeae confront clinicians worldwide with triple resistance—to penicillins, tetracyclines and fluoroquinolones38,39. Because of the need to provide a single-dose therapy to this highly noncompliant population of infected individuals, a parenteral cephalosporin is the only treatment remaining. A recent report of decreased susceptibility to cefixime forewarns the future demise of this last-resort family of drugs for gonorrhea40.

Today, MRSA strains that differ from the hospital strains and possess a new virulence toxin (Panton-Valentine leukocidin) have emerged in communities of industrialized countries41,42. The so-called 'community-acquired MRSA' is resistant to the β-lactam antibiotics, requiring physicians to commence alternative therapies when MRSA is suspected. Children were found to succumb to community-acquired MRSA infection because the disease had become too far advanced by the time that another effective therapy was initiated43. M. tuberculosis, particularly in some endemic areas, bears resistance to as many as eight drugs, making some individuals with tuberculosis incurable14. Previously (inadequately) treated individuals are at greatest risk; in some areas, more than 50% of such individuals have MDR tuberculosis (Box 2).

The frequency of drug resistance in the community has extended the resistance problem beyond the confines of the hospital. Resistant strains can be traced from the community to the hospital and vice versa, indicating that drug resistance is no longer localized.

What causes drug resistance?

The resistance problem can be seen simplistically as an equation with two main components: the antibiotic or antimicrobial drug, which inhibits susceptible organisms and selects the resistant ones; and the genetic resistance determinant in microorganisms selected by the antimicrobial drug44,45. Drug resistance emerges only when the two components come together in an environment or host, which can lead to a clinical problem. Selected resistance genes and their hosts spread and propagate under continued antimicrobial selection to amplify and extend the problem to other hosts and other geographic locations. There are more than 15 classes of antibiotics1 whose targets are involved in essential physiological or metabolic functions of the bacterial cell (Table 1). None has escaped a resistance mechanism1. Millions of kilograms of antimicrobials are used each year in the prophylaxis and treatment of people, animals and agriculture globally1,46,47,48, driving the resistance problem by killing susceptible strains and selecting those that are resistant.

Table 1 Major antibiotic families and their mechanisms of action

But how do bacteria acquire resistance? Drug resistance is mobile—the genes for resistance traits can be transferred among bacteria of different taxonomic and ecological groups by means of mobile genetic elements such as bacteriophages, plasmids, naked DNA or transposons1,49 (Box 3). These genes are generally directed against a single family or type of antibiotic, although multiple genes, each bearing a single drug resistance trait, can accumulate in the same organism. And, like the antibiotics themselves, resistance mechanisms are varied (Box 4).

In the absence of plasmids and transposons (which generally mediate high-level resistance), a step-wise progression from low-level to high-level resistance occurs in bacteria through sequential mutations in chromosomes1,33,50. This process was responsible for the initial emergence of penicillin and tetracycline resistance in N. gonorrhoeae. The organism later acquired transposons bearing genes with high-level resistance to these drugs. Strains of E. coli and other Enterobacteriaceae have evolved increasing resistance to fluoroquinolones, the result of mutations in the target enzymes (topoisomerases) and an increase in the expression of membrane proteins that pump the drugs out of the cell33,50,51.

Chromosomal mutants of S. aureus bearing intermediate resistance to vancomycin first appeared in response to vancomycin use20, only to be followed by those that acquired the high-level resistance transposon from enterococci22,23. A small increase in the minimum inhibitory concentration to an antimicrobial should alert clinical microbiologists in hospitals and communities to an incipient problem of resistance. Although still classified as 'susceptible,' a strain with decreased susceptibility to a drug heralds the eventual emergence of higher-level resistance and should galvanize efforts towards altering the use of that antimicrobial in that environment.

Resistant bacteria accumulate multiple resistance determinants

The long-term use of a single antibiotic (that is, for more than 10 days) will select for bacteria that are resistant not only to that antibiotic, but to several others1,52. This phenomenon was found to occur after the prolonged use of tetracycline for urinary tract infections53 and for acne54. Under continued antimicrobial selection, the susceptible intestinal and/or skin flora may become colonized by organisms that are resistant not only to the ingested drug, but also to other, structurally unrelated drugs. In animals, MDR emerged after the application of subtherapeutic (growth promotion) levels of tetracyclines in feed55. Within days, chickens began excreting tetracycline-resistant E. coli; by two weeks, the excreted E. coli were resistant to several antibiotics.

This phenomenon reflects the linkage of different resistance genes on the same transposon or plasmid. It is unclear, however, why multiple resistance plasmids eventually emerge with the prolonged use of a single antimicrobial. Bacteria that are already resistant to one growth-inhibitory agent seem to be favored in recruiting additional resistance traits from other bacteria sharing the environment: it was from the doubly resistant (penicillin and tetracycline) strains of N. gonorrhoeae that the new fluoroquinolone-resistant strains emerged.

Loss of resistance is slow

Resistant bacteria may rapidly appear in the host or environment after antibiotic use, but they are slow to be lost, even in the absence of the selecting antibiotic. This phenomenon reflects the minimal survival cost to the emerging resistant strains. In addition, as discussed above, resistance genes are often linked with genes specifying resistance to other antimicrobials or toxic substances on the same plasmids56. The presence of MDR plasmids assures maintenance of the plasmid as long as any one of the resistances provides a survival advantage to the host bacterium. This principle also applies to determinants of resistance to biocides such as quaternary ammonium compounds, because biocide efflux genes can be found on plasmids bearing genes for resistance to antibiotics in S. aureus57.

Some studies have, however, tracked a decline in resistance frequencies when an antibiotic is removed58. A significant countrywide reversal of macrolide resistance in S. pyogenes resulted from a Finnish nationwide campaign to reduce macrolide usage. In 2 years, resistance declined from about 20% to less than 10%59. Nonetheless, resistance generally persists at some low level and reintroduction of the antimicrobial will reselect resistant strains despite months or even years of nonuse.

Replacement by susceptible flora represents a chief contribution to a decrease in resistant strains. For example, despite being put into clean cages, chickens previously fed tetracycline-laced feed were found to continue to excrete tetracycline-resistant E. coli at high frequencies. When placed in separate cages and moved to a new location in the barn every 2–3 days, however, the resistance frequency dropped60. This 'dilution' of resistant strains was similarly accomplished by housing the chickens with greater numbers of cage mates that excreted susceptible flora. The findings suggest that the fastest way to eliminate resistant strains is to outnumber them with susceptible strains.

The ecology of antibiotic resistance

The impact of the drug selection process can be largely confined to the individual taking the antibiotic if widespread antibiotic usage is absent. After therapy, the selected resistant commensal strains will eventually be 'diluted out' and their growth will be suppressed by the return of drug-susceptible, natural competitors. If, however, whole populations are being treated with the same class of antibiotic, susceptible strains will have little opportunity to recolonize their niche and resistant strains will acquire an important advantage. The resulting ecological imbalance produces a potentially serious environmental pool of resistance genes61.

Ecologically speaking, it is the density of antibiotic usage that enhances resistance selection and its effects. The 'selection density' involves the total amount of antibiotic being applied to a geographically defined number of individuals in a setting, whether it is the home, daycare center, hospital or farm62 (Fig. 1). Each individual becomes a 'factory' of resistant bacteria that enter the environment. The disparity between resistance rates in the local community and those in city hospitals reflects differential ecological effects of antibiotic use. The end result of the selective pressure will reflect the number of individuals who are contributing resistant bacteria to that environment and the residual number of surviving, susceptible bacteria.

Figure 1: Post-therapeutic effects of antibiotic dispersion.
figure 1

(a) While on antimicrobial therapy, the individual (e.g. person or animal) is a focal point for a high concentration of both antibiotic (red shading) and resistant bacteria (black dots) that are selected and generated from its use. (b) Over time, resistant bacteria spread to local contacts and antibiotic enters the environment through waste and water disposal (for example, from animals) or sewage (from people). If several individuals are treated, however (c), a higher density of antimicrobial and resistant organisms is established in the same environment (d). The selective process is continuous, occurring both during and after therapy.

The ecological effects of antibiotics make them unique therapeutic agents. They are 'societal drugs' in which individual use affects others sharing that environment62,63. For example, antibiotic treatment for acne was found to produce an MDR skin flora not only in the individual with acne, but also in other members of the household64. High numbers of MDR bacteria were found in the intestinal flora of ambulatory individuals in the Boston area, even though none had recently taken an antibiotic65. In Nepal, resistance rates in individuals were found to correlate more with the total community use of antibiotics than with the individual's own use66.

In addition, the selection of resistance continues because antimicrobials persist, largely intact, in natural environments. Antimicrobials in waste waters are being reported with increasing frequency and are potentially important contributors to the environmental selection of antibiotic-resistant organisms67. The findings suggest that one approach to the antibiotic resistance problem could be to design drugs that self-destruct after treatment, thereby removing a contributing factor in the propagation of resistance.

Use of antibiotics in food animals and agriculture

Considerable debate surrounds the relationship between antimicrobial use in animals and the resistance problem in people47. The chronic use of subtherapeutic amounts of antibiotics for growth promotion in food animals has been banned in the European Union, but it continues in the United States, albeit under intense scrutiny by the Center for Veterinary Medicine of the Food and Drug Administration. Despite their low-level application, the antibiotics select determinants mediating high-level, clinically relevant resistance55. Enteric organisms such as Salmonella, Campylobacter, Listeria, enterococci and some strains of E. coli are propagated primarily among animals and subsequently infect people. The transfer may occur through the food chain or through animal handlers68,69,70,71. If the organisms are MDR, the emergence of their resistance results principally from use and overuse of antibiotics in the animals. Overall, animal contributions to the resistance problem in human infections are small but not insignificant; they have a major role if enteric organisms are involved.

Antibiotics also enter the environment through the dusting of fruit trees for disease prophylaxis72 and the application of antibiotic-laden animal manure on croplands1,47. These varied applications all add to the continued selection of resistant bacteria.

How can we manage and prevent drug resistance?

Track the resistance frequency. Local, national and global surveillance systems of drug susceptibility would help to communicate the current status of resistance in a location, facilitating more appropriate choices of treatment. Such surveillance would alert public health officials to new pathogens and would spur the implementation of control policies. In this regard, the Alliance for the Prudent Use of Antibiotics has established its Global Advisory on Antibiotic Resistance Data project to synthesize, evaluate and report the surveillance data from five large global surveillance systems (Box 2).

Commensal organisms are common reservoirs of antibiotic resistance plasmids, transposons and genes. E. coli and the enterococci of the gut serve as reservoirs from which several antibiotic resistance genes can spread73. The commensal Haemophilus parainfluenzae has been shown to confer β-lactamase-specifying plasmids to H. influenzae52. Similarly, Staphylococcus epidermidis serves as a reservoir for resistance genes and plasmids for the more pathogenic S. aureus52,74. Vancomycin resistance determinants found initially among enterococci appeared in other commensal bacteria before emerging in S. aureus9. This concept has been recently formalized by an Alliance for the Prudent Use of Antibiotics–based Reservoirs of Antibiotic Resistance project that supports studies examining the link between resistance in commensal flora and resistance in clinical isolates (Box 2).

Isolate hospitalized individuals with potentially dangerous resistant bacteria: cohorting. In Perth, Australia, hospital patients colonized with MRSA are isolated in special units, a process that has led to the lowest levels of MRSA and MDR staphylococci among all Australian hospitals75,76. Similar measures in Scandinavian countries and Holland protect hospitals from the entry and spread of resistant, difficult-to-treat infectious disease agents. In the United States, individuals with MRSA or vancomycin-resistant entercocci are housed in single rooms and kept microbiologically isolated. Although more than this single measure is needed to reduce the spread of MRSA, a review of this practice has identified six well-designed studies with a positive outcome and has concluded that cohorting should be “continued until further research establishes otherwise”77.

Introduce new therapeutic approaches. Confronted with a shortage of new antimicrobials, we must use our current drugs more prudently. Reducing and improving use can diminish resistance and permit a drug to resurface eventually as an effective therapy58. The appropriate use of the antibiotics not only can help to reverse high resistance frequencies, but also can curb the appearance of resistance to newer agents58. Decreasing antibiotic usage in the intensive care and other hospital units has shown that susceptible indigenous strains will repopulate the ecological niche in the absence of drug-selective pressure. But the process is slow and more difficult when addressing MDR strains, for which the use of many antibiotics must change to affect the presence of that strain. In addition, such efforts cannot succeed alone78. They need to be complemented by other actions (see below). For tuberculosis, better compliance after 'directly observed therapy' has clearly proved to be effective in treating the disease and in preventing the emergence of resistance79. Continued use of the same drugs in areas where resistance is endemic should be halted. From what we have learned, shorter-course therapies with highly active antibiotics will also reduce the pressure on multidrug resistance.

The development of new antibiotics—either those that block or circumvent resistance mechanisms or those that attack new targets—is essential. Such antibiotics would evade current resistance mechanisms, which can thwart the success of new, but structurally similar drugs. A different approach focuses on preventing infection by inhibiting key gene products that are involved in the infection process itself80. Because the inhibition of these targets does not affect growth, selection for resistance should be considerably reduced. The pipeline for new drugs is small, because the major pharmaceutical companies have largely abandoned the antibiotic discovery field81. Fortunately, the need is being addressed by small, often start-up companies that can devote full attention to this goal but will ultimately require support from investors or from the larger pharmaceutical industry.

The availability of rapid diagnostics for the healthcare provider would greatly enhance the ability to prescribe more appropriately. A test to distinguish a viral from a bacterial infection, for example, one based on procalcitonin levels82, should decrease unnecessary antimicrobial use. More rapid susceptibility tests would aid the initial selection of an antibiotic. There is no better need for such diagnostics than for early stage tuberculosis, before the foci of resistant strains can spread out of control.

Finally, the development of conjugated vaccines, such as those based on encapsulated H. influenzae type b and pneumococcus, can diminish bacterial disease and the consequent need for antibiotics. But vaccine development and delivery are problematic. In addition, with the transformation of commensal strains into pathogens in immunocompromised individuals, the activity of vaccines against these organisms could paradoxically destroy a natural defense against recognized pathogens.


The erosion of effective antimicrobials continues as we witness the increased frequency of resistance to all drugs—in particular, the fluoroquinolones, vancomycin and carbapenems, which are often the drugs of last resort. Imminent crises have been averted by new drugs that can combat MDR Gram-positive bacteria. With the relative absence of new antimicrobials coming to market and with new threats arising from the Gram-negative bacteria, however, the number of drug options leaves us perilously close to none or only a single effective agent for some life-threatening infections.

Hundreds of β-lactam-degrading enzymes are rapidly undermining the mainstay penicillins and late-generation cephalosporin agents. The increase in metallo-β-lactamases, which are active against carbapenems and most other β-lactams, is an alarming new development32. Colistin, a relatively toxic drug, has become a last-resort choice in treating some strains of P. aeruginosa83. In addition, new types of highly virulent MRSA in the community are posing concerns for everyday activities among populations at risk, including children, contact sports participants, the military and economically deprived indigenous populations41,42. Notably, organisms that were formerly classified as primarily 'commensal', namely enterococci, pneumococci and E. coli, as well as environmental organisms such as P. aeruginosa and A. baumanii, have become emerging pathogens. The narrow focus on the older clinical pathogens must be broadened to accommodate the trend toward these newer disease agents, which are largely panresistant. From an ecological perspective, contamination of the environment with antibiotics from human, animal and agricultural spillover continues to exert selective pressure for resistance determinants.

Improved technologies have identified the clonal nature of infectious agents, enabling us to track their movement more closely and to understand better their epidemiology. Notable examples are MDR tuberculosis84, Streptococcus pneumoniae85 and some strains of cotrimoxazole-resistant E. coli86. Such advances have facilitated a broader view of resistance as an ecological problem. Few, if any, barriers are able to contain resistance genes and their bacterial hosts in our closely connected world.

Although drug resistance has been recognized since the early 1940s, and despite many national and international reports, including that of the World Health Organization87, urging ways to curtail it, the problem continues to grow and to evolve from one decade into the next. The highly disease-oriented focus of modern medicine has hindered a clear perception of the enormity and all-encompassing nature of resistance, which suffers from an 'identity crisis.' Resistance is a nameless cloud that looms over otherwise controllable infections, but lacks the powerful status of a readily identifiable disease state to spur large-scale efforts of control.

The role of antibiotics in the treatment of infectious diseases cannot be seen as anything but essential for the foreseeable future. The obstacles of few new antimicrobials on the horizon and the increasing frequency of multidrug resistance mean that we must redouble our efforts to preserve the agents at hand, while intensifying the search for new therapeutics.