Introduction

Pseudomonas aeruginosa is one of the top nosocomial pathogens and thus a leading cause of severe hospital-acquired infections, particularly in critically ill and immunocompromised patients [1]. As a proof of its clinical relevance, for instance P. aeruginosa is responsible for approximately 4.7–8.9% of hospital-acquired bloodstream infections and multidrug resistant (MDR) P. aeruginosa responsible of 26.3% of MDR bacteremia [2,3,4],with mortality rates of around 30% [5,6,7].

Different studies have shown that P. aeruginosa infections are associated with worse clinical outcomes than other pathogens [8,9,10,11,12,13,14,15]. To give just one example, Thaden et al. evaluated the impact of bacterial species (Staphylococcus aureus or gram-negative bacteria) on outcome in bloodstream infections in a 13-year prospective study involving more than 2,600 hospitalized patients with bloodstream infection and found a significant association between P. aeruginosa and higher in-hospital mortality compared to other bacteria after adjusting for patient demographics, medical comorbidities, bacterial antibiotic resistance, and treatment factors [16].

A key aspect of the success of P. aeruginosa as an opportunistic pathogen is its extensive genomic complexity and metabolic plasticity, which includes several intrinsic defense mechanisms, an extensive array of virulence genes and biofilm production capacity [17]. Connected to this is the extraordinary capacity of this species to develop and spread antimicrobial resistance in vivo [18, 19].

The worldwide emergence of antimicrobial resistance has become a major public health concern [20]. P. aeruginosa is one of the top six pathogens responsible for resistance-associated deaths [20] and is listed among the ESKAPE pathogens [21]. The WHO lists carbapenem-resistant (CR) P. aeruginosa among the ‘critical priority’ pathogens for which new antibiotics are urgently needed [22].

Regarding P. aeruginosa epidemiology, in a recent study conducted from 2018 to 2022 by the ATLAS (Antimicrobial Testing Leadership and Surveillance) global program, the rates of carbapenem-resistant P. aeruginosa varied from 15 to 33% depending on the region. Difficult-to-Treat Resistant P. aeruginosa rates varied from 6% in North America to 12% in Latin America in 2022 and remained stable over time in all global regions, except for a signifcant decreasing trend in isolates from Europe [23]. In the European Union/European Economic Area, the European Center for Disease Control data for 2023–2021 noted considerable differences in the percentages of CR strains, ranging from under 5% in two of 44 reporting countries to equal to or above 50% in 6 countries [24]. As recently as the first decade of this century, the absence of standardized international definitions of the degrees of multidrug resistance made it difficult to contrast the results of different studies, [25]. In 2012, a consensus document was published proposing international definitions of the terms multidrug resistance (MDR) (non-susceptible to ≥ 1 agent in ≥ 3 antimicrobial categories), extensive drug resistance (XDR) (non-susceptible to ≥ 1 agent in all but ≤ 2 categories) and pandrug resistance (PDR) (non-susceptible to all antimicrobial agents) [26]. However, in the case of P. aeruginosa, this proposal has certain limitations, since it was validated before the introduction of the novel β-lactam–β-lactamase inhibitor (BL-BLI) combinations, and the results vary depending on breakpoints used (The European Committee on Antimicrobial Susceptibility Testing—EUCAST or The Clinical & Laboratory Standards Institute- CLSI). Furthermore, the definition of XDR was formulated in terms of the number of antimicrobial categories with resistance but did not differentiate between first and second or third-line antimicrobial agents. In 2018, the IDSA (Infectious Diseases Society of America) proposed a new definition for pathogens with non-susceptibility to all first-line agents called Difficult-to-Treat Resistance (DTR), defined as resistance to all first-line (high efficacy and low toxicity) agents, which, in the case of gram-negative bacteria, meant resistance to fluoroquinolones and to all ß-lactams (including carbapenems) regardless of the susceptibility to the BL-BLI combinations and aminoglycosides [27]. Finally, a recent update by the European Study Group for Antimicrobial Resistance Surveillance of the European Society of Clinical Microbiology and Infectious Diseases suggests including 5 new categories, considering that neither the European Centre for Disease Prevention and Control nor the IDSA definitions take into consideration the newly approved β-lactams and BL-BLI combinations: ceftolozane/tazobactam, ceftazidime/avibactam, imipenem/relebactam, meropenem/vaborbactam and cefiderocol [28]. In the present review, we use as reference the term DTR P. aeruginosa. However, we maintain the original definitions of the studies named in the manuscript (CR, MDR and XDR) for a better interpretation of the outcomes.

The aim of the present research is to review the impact of multidrug resistance on bacterial virulence and fitness of P. aeruginosa. We discuss the microbiological implications, consequences for clinical outcomes and the economic burden of resistance in P. aeruginosa.

We try to make an integrative review of the resistance-virulence balance from all perspectives, from the basic and molecular research, through topics of molecular epidemiology and implications in large collections of clinical strains, until reaching patient cohorts and outcomes for them. To our knowldege there is no similar review in the literature, which allows to understand all the implications of resistance from the beginning (at the level of bacteria isolated in vitro) to the end (patient).

Virulence factors in Pseudomonas aeruginosa

The virulence of a pathogen is its ability to infect the host and cause clinical symptoms, assisted by factors that promote bacterial adhesion, colonization, invasion of the host tissues, dissemination as well as evasion of the host immune response [31]. P. aeruginosa is endowed with a wide array of virulence factors that contribute to pathogenesis in acute and chronic infections [32], and regulated by complex, interconnected pathways and signaling systems that confer great plasticity to this pathogen [33]. Some of these virulence determinants are in the accessory genome acquired by horizontal gene transfer, which represents at least 10% of the genome, forming part of pathogenicity (PAPI) islands, such as PAPI-1 and PAPI-2 (responsible in part for the hypervirulence of strain PA14) or genomic (PAGI) islands [34,35,36,37].

Virulence determinants include cell-associated and extracellular factors. A list of the most important ones in P. aeruginosa is shown in Table 1.

Table 1 Main virulence determinants of Pseudomonas aeruginosa
Full size table

Regarding cell-associated factors, on the bacterial surface there are appendages, such as flagella and type IV pili, which are the main cell adhesion and motility systems; and exopolysaccharides, also responsible for adhesion to host cells and biofilm formation (including the lipopolysaccharides: alginate and Pel and Psl).

Lipopolysaccharide is a major virulence factor present in many gram-negatives, comprised of lipid A (which anchors the structure in the outer membrane), core oligosaccharide, and O-antigen, which is the most distal element. This factor is a major structural component of the outer membrane, enabling resistance to the complement system, several antimicrobial peptides and toxic compounds, and exhibiting important pro-inflammatory effects. The O antigen has also been used to classify P. aeruginosa isolates [36,37,38] and certain variants are associated with a worse prognosis in respiratory tract (O1 and O11 serotypes) [40, 41] and bloodstream infections (O11 serotype) [29, 42].

Among the extracelular factors are included toxins, elastases, pigments, siderophores (pyoverdine and pyochelin) and other proteases. One of the most important P. aeruginosa virulence determinants is the type III secretion system (T3SS), which injects effector cytotoxins into host cells (ExoS, ExoT, ExoU and ExoY) [43, 44]. These exotoxins disrupt the epithelial barrier and so inhibit cell migration and proliferation, inducing apoptosis [45]. Of the four effector exotoxins identified, ExoU is the most potent as it leads to rapid necrosis of the epithelial barrier [44] and its expression has been associated with higher bacterial virulence and worse clinical outcomes in patients with pneumonia and bacteremia in both clinical and experimental studies [5, 38, 42, 46, 47]. The siderophores (pyoverdine and pyochelin) are produced by the bacterium as iron acquisition systems, vital for its growth and virulence [33, 48,49,50,51].

Exotoxin A is an ADP-ribosyl transferase that inhibits protein synthesis, inducing cell death, cytokine production, and reducing the host’s response to infection. LasA and LasB proteases exhibit elastolytic activity, apromoting degradation of blood vessel and pulmonary alveoli tissues. The alkaline protease and protease IV degrade complement proteins, cytokines (alkaline protease) and host clotting factors (protease IV). The rhamnolipids contribute to the pathogenesis of P. aeruginosa lung infections by degrading alveolar surfactant and participating in ciliostasis [33, 52,53,54].

Finally, quorum sensing, also secreted to the extracellular medium for bacterial cell-to-cell communication, involves the production, detection, and response to signaling molecules called autoinducers, that regulate gene expression, including a myriad of genes responsible for virulence-related behaviour. Thus, this system plays a key role in bacterial virulence, resistance, and biofilm formation [32].

Pseudomonas aeruginosa epidemic high-risk clones

Pseudomonas aeruginosa shows significant clonal diversity, with more than 4,000 multi-locus sequence typing (MLST) profiles described (http://pubmlst.org/paeruginosa) [55]. However, this diversity primarily affects susceptible strains showing a polyclonal non-endemic pattern [56]. In contrast, most P. aeruginosa strains with a MDR/DTR phenotype, belong to a few endemic clones, known as high-risk clones [29, 30]. Multiple reports in recent decades have reported the worldwide spread of these clones [57,58,59], with clones ST235, ST111, and ST175 being the most prevalent globally [29]. These high-risk epidemic clones are characterized by their exceptional ability to acquire resistance determinants through horizontal gene transfer, especially horizontally-acquired-β-lactamases such as extended spectrum β-lactamases (ESBLs) and carbapenemases [28, 29, 60]. Table 2, adapted from del Barrio-Tofiño et al. [60], shows the main virulence characteristics, geographic distribution, and association with the main horizontally-acquired β-lactamases of the most prevalent high-risk clones.

Table 2 Characteristics of the main international Pseudomonas aeruginosa high-risk clones
Full size table

ST235 (O-antigen serotype O11) is considered the most significant high-risk clone with a wide distribution worldwide [29, 58] and is associated with over 60 different β-lactamase variants, including numerous Class A and B carbapenemases. ST111 (serotype O12) has also been documented on every continent except Oceania and is usually associated with the presence of metallo-β-lactamase (MBL) VIM-2. ST175 (serotype O4), is widely distributed in several European countries and Japan, and has a specific chromosomal mutation in AmpR (G154R) that causes hyperproduction of the intrinsic β-lactamase AmpC, OprD (Q142X), and MexZ (G195E), and 3 quinolone resistance-determining region (QRDR) mutations (GyrA T83I, D87N and ParC S87W) [58, 60,61,62].

Other clones less extended are summarized in Table 2.

Different studies have evaluated the pathogenicity [60] and virulence of these high-risk clones [5, 64, 65]. Peña et al. determined that the T3SS genotype and high-risk clones are significantly linked, with specific high-risk clones showing well-defined T3SS genotypes [5]. Clone ST235 is known to have a greater capacity to acquire antibiotic resistance determinants and increased virulence [5, 66]. It is related with the exoU + virulence factor and the O11 serotype [28, 58, 67]. This clone has also been shown to be associated with increased mortality in bloodstream infections compared to other high-risk clones [47, 68]. On the other hand, other prevalent clones, such as ST111 and ST175, show the exoU-genotype and thus lower cytotoxicity, invasiveness, and virulence [62]. Recio et al. found that ST175 strains were associated with particularly low virulence, while the pathogenicity of ST235 strains was similar to that of non-MDR strains, despite exhibiting a wide range of acquired resistance mechanisms. Their findings also lend support to the T3SS genotype as one of the most important virulence determinants, since of the high-risk clones studied, ST235 was the only one with the exoU+ [69]. A Spanish multicenter study by Mulet et al., found that the three major high-risk clones (ST111, ST175, and ST235) were associated with low levels of virulence markers such as motility and pigment production, as well as reduced bacterial fitness, but also with increased spontaneous mutation frequency and biofilm growth, likely enabling a greater capacity for dissemination [70]. Other studies have demonstrated increased biofilm formation, high levels of pyocyanin and decreased motility in high-risk clones [71, 72].

As for the clones typically associated with cystic fibrosis, the most important and the first to be described was the Liverpool epidemic strain, or Clone C [73]. This clone has traditionally been considered non-virulent and differs by only 1 nucleotide from clone ST598 [74, 75]. Another clone widely distributed in patients with cystic fibrosis is ST274 (belonging to the clonal complex, CC274) [76]. This clone is associated with the presence of β-lactamases OXA-486 and PDC-24 genes, the O3 serotype and the exoU- genotype [77].

Balance between antibiotic resistance and fitness/virulence in Pseudomonas aeruginosa: fundamental concepts and basic research

Fitness (also referred to as biological success) is the capacity of bacteria to survive and multiply within a specific niche and has traditionally been measured through growth rate and interbacterial competitive assays. When considering a pathogenic species such as P. aeruginosa causing acute infection, fitness is invariably linked to virulence, since the greater the capacity to infect and disseminate thanks to its virulence factors, the more likely it is to thrive inside the host, and vice versa [78, 79]. Given this close relationship, here we shall use fitness and virulence interchangeably, and the terms biological cost (or burden) and virulence attenuation as synonyms although we know that strictly speaking, they are not. Therefore, we will use them in a sensu lato conception to simplify the complex connection with resistance and make the reading easier. In other words, both if a bacterium grows slower in vitro or if it shows a very specific toxin hypo-expression for instance, the general outcome would be similar: an infection less harmful for the host. In any case, when in this section the consequences for biological success linked to each resistance mechanism are approached, we tried to deeply dissect them, clearly stating if attenuation is due to a general fitness impairment or linked to specific downregulations of pathogenesis-related features, which substracts importance of our indistinct use of concepts. However, it must be clarified that the analysis of the interplay between resistance and fitness/virulence we make here is referred to the acute infection context, because in the chronic infection (e.g. cystic fibrosis), this interplay is much more complex and particular. For instance, in this chronic scenario certain features such as slow growth and pathogenic attenuation are often positively selected, although these traits are deleterious if applied to acute infection [80]. Thus, the balance between antibiotic resistance and biological success in chronic respiratory infections is a very particular topic that would deserve an especially dedicated review and will not be considered in this manuscript.

Although it has been accepted for years that there is inevitably some fitness cost associated with acquired resistance in the context of acute infection, this assumption is not so clear-cut in reality [79, 81,82,83]. The view is based on the fact that an acquired mutation or horizontal determinant often has negative side effects for bacterial biology, such as alterations in antibiotic targets leading to decreased protein efficiency, increased energetic costs associated with the (hyper)production of enzymes that inactivate the antibiotic, unwanted extrusion of certain useful compounds in addition to the drug, reduced nutrient uptake because of porin loss, etc. This assumption justifies the research interest in looking closely at resistance and fitness with the ultimate goal of discovering therapeutic targets to reduce the pathogenic power of bacteria and/or hinder the emergence of resistance by taking advantage of the potentially high associated biological costs [30, 84]. However, as will be seen below, there are multiple examples of cases where there appears to be no biological burden, either because the acquired mechanisms themselves appear to be cost-free, or compensatory mutations are selected, or even because other markers of increased virulence are co-carried with the acquired resistance gene on mobile genetic elements [81, 85, 86]. There are several studies with clinical strains or mutants evolved in vitro, in which the resistance phenotype is treated as the end result of a combination of mechanisms, but it is very difficult to establish direct cause-effect relationships between specific resistance pathways and biological costs in these studies. In this section on the other hand, we review the resistance/fitness trade-off from a molecular perspective, collecting data in which different resistance mechanisms and their respective biological burdens are addressed separately. Thus, here we take a close look at the main knowledge in the field of basic research on the connection between resistance and fitness/virulence in P. aeruginosa, dividing the topic into horizontally-acquired versus mutation-driven resistance for ease of understanding. Subsequently, the clinical perspective of the interplay between resistance and virulence in P. aeruginosa will also be addressed to provide an integrated overview of the topic.

Horizontally-acquired resistance

It has to be considered that the biological cost associated to horizontal resistance genes may be often due to the the carrier elements (usually plasmids) themselves rather than to the specific resistance determinant. However, since multiple factors could determine the cost of each particular plasmid [87,88,89] (size, copy number, codified genes, or even compensatory features such as pathogenicity islands, etc.), to simplify the approach in this section we dissect the knowledge about the biological burden, apparently specific, associated to the main horizontally-acquired determinants.

The paradigmatic example of horizontally-acquired resistance is the incorporation of β-lactamases—typically encoded by genes in class I integrons—carried on mobile elements such as plasmids, although few studies have specifically analyzed their associated biological costs. The results in this regard are variable, ranging from β-lactamases for which no associated attenuation was observed to others manifesting significant measurable impairment of fitness/virulence linked to certain enzymes. Among the former, some examples can be cited, such as the class B carbapenemases IMP-1 and VIM-1, the class A penicillinase PSE-1, the class A extended-spectrum β-lactamase [ESBL] GES-1, the class C cephamycinases FOX-4 and FOX-8, and the narrow spectrum class D OXA-2 enzyme [90,91,92] Conversely, other β-lactamases have been shown to entail important burdens, such as the case of class A TEM-1 and class D OXA-3 enzyme production, which triggered impaired twitching motility, adhesion, and biofilm formation by P. aeruginosa without affecting its growth [93]. More recently, production of the OXA-2-derived ESBLs, OXA-226, OXA-161 and especially OXA-539, was shown to be associated with significantly attenuated virulence in an invertebrate model [92]. Although the underlying basis for these high burdens has not been fully determined, some possible explanations have been advanced. For instance, given the likely common origin with ancestral enzymes showing peptidoglycan-lytic activity, it was proposed that some β–lactamase variants could still display certain capacity to metabolize the own cell wall nowadays [87, 94]. Thus, in addition to hydrolyzing the β–lactam drug, in certain circumstances these β-lactamases could be weakening the cell wall by partially degrading it, which would impair cell viability, obviously entailing a great biological burden. Moreover, this residual activity could affect the peptidoglycan-derived molecules, altering the pool of soluble fragments and triggering a downregulation of pathogenesis through appropriate regulators, similar to what happens with AmpR and the muropeptide-linked modulation of AmpC β-lactamase expression [95]. The fact that a functional active site of TEM-1 was required to cause the abovementioned biological burden associated to its production [93], would support the idea of a residual peptidoglycan-degrading activity being the key for this outcome, although other explanations have not been ruled out.

From the abovementioned data, it can be concluded that most of the relevant horizontally-acquired β–lactamases have not yet been analyzed from the perspective of their impact on fitness/virulence. However, considering the worldwide epidemic dissemination of certain enzymes (including OXA-10, PER-1, NDM-1, KPC-2, VIM-2, certain GES-type ESBLs/carbapenemases, among many others), which are often associated with successful high-risk clones, it may be assumed that their costs must be low overall if these associated costs were high, these enzymes would not be so globally widespread [28, 60]. By the same token, a very high cost linked to specific amino acid variants (the aforementioned OXA-539 for instance) [92] would explain their lack of epidemiological success despite the increased capacity for β–lactam hydrolysis. In any case, the possibility of some connections between certain high-risk clones and quite specific β–lactamases being caused by particular traits of the strain but not by the absence of biological cost associated to each β–lactamase per se has not been ruled out. Thus, the opposite phenomenon cannot be discarded either, i.e. that some β-lactamase variants such as the abovementioned OXA-2 derived ESBLs could globally disseminate in future if they get to be acquired by high-risk clones in which the fitness costs of these enzymes may be minor. Although this interesting topic has been approached in Escherichia coli [96], and was shown to explain why some β-lactamases are seemingly limited to certain species and absent in others (e.g. VIM-2 and SPM-1 in Acinetobacter baumannii [97]) remains to be investigated in P. aeruginosa, posing an interesting field worth to be opened.

Athough besides β-lactamases the acquisition of aminoglycoside-modifying enzymes in class I integrons is a typical hallmark of P. aeruginosa clinical strains [98], the topic of their potential biological costs has been never investigated in this species. Therefore, although a presumable low burden can be deduced (if it was high, these enzymes would not be that disseminated) whether these determinants entail different costs depending on the family (N-acetyltransferases, O-nucleotidyltransferases and O-phosphotransferases) and/or the host strain are questions that remain to be approached in future. Conversely, recent work has studied the biological cost associated with the horizontally-acquired colistin resistance determinant, Mcr-1 (in which the addition of phosphoethalonamine to lipopolysaccharide reducing colistin affinity for its target), revealing a lack of impact on fitness/virulence in P. aeruginosa [99]. This represents a red flag for the potential dissemination of this determinant in P. aeruginosa (at the moment, this mechanism is much more prevalent in Enterobacteriaceae), threatening the future antipseudomonal effectiveness of this drug of last resort.

Mutation-driven resistance

With respect to mutation-driven resistance, probably the most important mechanism in P. aeruginosa is hyperproduction of chromosomal AmpC cephalosporinases, which affects 3rd- and 4th-generation cephalosporins, as well as antipseudomonal penicillins and monobactams. Different mutations in genes related to peptidoglycan metabolism cause this hyperproduction [48], with no demonstrable biological burden associated with the two most common pathways, namely, disruption of the ampD and dacB genes [100]. On the other hand, when AmpC hyperproduction was combined with presence of a PSE-1 enzyme in a murine infection model, dramatic attenuation was reported [91], suggesting that AmpC hyperproduction in itself is a very potent, cost-free resistance mechanism and may have deleterious effects only when combined with certain other circumstances [91, 100]. In this context, it was recently shown that the burden of AmpC hyperproduction occurring only in very particular scenarios (peptidoglycan recycling blockade) is dependent on enzymatic activity and reinforces the idea that the abovementioned residual peptidoglycanase activity in certain β-lactamases may be a key contributor to impaired fitness/virulence [92, 100,101,102]. In fact, this need for an intact active site had been already shown for TEM-1 [93], supporting the idea of a partial degradation of the own murein sacculus due to this residual activity, as the cause of cell viability impairment and the necessarily derived fitness cost. Obviously, blocking an essential pipeline for the production of new material to be incorporated into nascent peptidoglycan, as are the recycling routes, would further weaken the cell wall revealing or exaggerating the biological costs [100, 101]. These outcomes could be also caused by the residual peptidoglycanase activity acting on a presumptive pathogenesis regulatory muropeptide pool [95] and not only on the peptidoglycan as rigid structure, as abovementioned for certain horizontally acquired β-lactamases [92, 93].

As a result of the selective pressure exerted by the new drug combinations ceftolozane/tazobactam and ceftazidime/avibactam, there are increasing reports of certain amino acid variants appearing in chromosomal AmpC causing resistance to these antimicrobials when combined with hyperproduction [103]. Of these, only the T96I, G183D and ΔG229-E247 variants have been analyzed from a fitness/virulence perspective and have been shown not to increase costs relative to wildtype AmpC values [100]. In contrast, hyperproduction driven by an amino acid change (G154R) in the master AmpC regulator, AmpR, has been shown to contribute to reduced virulence in the Caenorhabditis elegans model, which is consistent with the fact that AmpR apparently governs the expression of certain virulence-related genes [104, 105]. However, taking into account that this AmpR variant is very closely linked to the high-risk clone ST175 (see "Pseudomonas aeruginosa epidemic high-risk clones" section), it is likely that the increased resistance to β–lactams that it affords more than compensates for the associated cost.

The mutation driven loss of the OprD porin, which confers increased resistance to carbapenems, has been associated with remarkably hypervirulent behavior, including increased cytotoxicity towards macrophages and lethality in mice, although the underlying basis for this phenomenon has not yet been resolved [106, 107]. Other studies of the same mechanism in murine models however have shown a slight reduction in fitness [91, 108], suggesting that the contradictions in these studies may be due to differences in the P. aeruginosa strains used and the fitness/virulence-related parameters analyzed.

Hyperexpression of efflux pumps is one of the most important resistance mechanisms in P. aeruginosa and affects practically all antipseudomonal drugs. The major efflux pumps in this context are MexAB-OprM, MexCD-OprJ, MexEF-OprN and MexXY-OprM, forming a complex modulatory system made up of different regulators with mutation profiles driving hyperproduction in each of them [109]. From the point of view of fitness/virulence, MexCD-OprJ hyperproduction, mediated by mutations in its nfxB repressor, seems to be clearly associated with attenuated virulence [110,11,112,113]. In the case of MexEF-OprN hyperexpression, deletion of this pump resulted in a corresponding increase in virulent behavior, with improved collagenase activity, rhamnolipid production and swarming motility, although the underlying basis of this hypervirulence is not fully understood [111, 114,115,116,117,118,119,120,121].

The information relating to MexAB-OprM is conflicting, with some data showing a significant biological burden [122, 123] contrasting with other results in which pump hyperexpression was associated with enhanced virulence [119, 124]. Of a similar contradictory nature, both MexXY-OprM hyperexpression and its deletion led to significant impaired virulence, including cell invasiveness and pyoverdine production, which once again demonstrates the complexity of the interplay between resistance and virulence [125, 126]. In general, attenuated virulence in the aforementioned pumps has been related to decreased release of extracellular factors such as proteases, elastase, pyocyanin and rhamnolipids (linked in turn to swarming motility) as well as impaired performance of the T3SS. A common explanation for the attenuated virulence often reported to be associated with hyperexpression of efflux pumps is that, in addition to extrusion of the antibiotics, the pumps are also expelling the quorum sensing signals (and thus decreasing their intracellular concentrations) needed for efficient production of different virulence factors [110,111,112,113,114,115,116, 118,119,120, 122, 123, 125].

DNA gyrase consists of two subunits encoded by the gyrA and gyrB genes, whereas the topoisomerase IV subunits are encoded by parC and parE. These proteins, which are involved in the DNA relaxation/coiling processes necessary for replication, are the targets of quinolones. Hence, certain amino acid changes in their sequences are associated with lower binding affinity of the drug for their targets and related resistance, which explains why they are known as Quinolone Resistance-Determining Regions (QRDR) [127]. The burden associated with certain QRDR mutations has been studied, and it has been shown that, in general, low levels of quinolone resistance are mostly cost-free, whereas the highest levels require sequential mutations in the mentioned genes, which obviously leads to increased biological costs. These have been linked to alterations in DNA supercoiling levels, consistent with impaired DNA gyrase and/or topoisomerase IV action [128]. However, it has been shown that the burden associated with QRDR mutations can be readily alleviated by compensatory mutations at different sites [83, 128]. Interestingly, it has been suggested that the T3SS toxin ExoU may better regulate DNA supercoiling, which relieves the burden associated with QRDR mutations [129], and would explain the reported correlation between the presence of this toxin and the high prevalence of QRDR mutations in exoU+ clinical strains [130].

In P. aeruginosa, the main mechanism of colistin (polymyxin E) resistance is the addition of 4-amino-4-deoxy-l-arabinose to lipopolysaccharide, which is associated with mutations in different two-component systems leading to overexpression of the arnBCADTEF operon, which encodes aminoarabinosylation enzymes [131]. It was recently shown that aminoarabinosylation in itself (achieved by cloning and constitutive hyperexpression of the enzymes involved) has a negligible impact on P. aeruginosa fitness and virulence [132]. However, it has also been shown that after 21 days of exposure to colistin in a morbidostat device, resistant mutants displayed other side effects such as improved serum resistance and biofilm formation, but decreased insect larvae killing capacity, demonstrating the complex implications of colistin resistance [133]. Closely related to colistin and in this same latter sense, it was shown that polymyxin B resistance development (whose mechanisms were not determined) was associated with a significant attenuation of P. aeruginosa virulence in the invertebrate model used, but this outcome was only seen in a DNA repair-deficient hypermutator strain (mutS KO mutant), not in the wildtype background [134].

A final example that contradicts the oversimplistic view that resistance usually carries biological costs is the mutational disruption of glpT (that encodes the transporter that enables fosfomycin uptake) which had no measurable effect on impaired fitness or even an increased pathogenic response, in parallel to fosfomycin resistance [106, 135, 136].

Balance between antibiotic resistance and fitness/virulence in Pseudomonas aeruginosa: collections of clinical strains and animal models

In the previous section we reviewed the interplay between resistance and biological burden from a molecular perspective, focusing on specific resistance mechanisms. In this section, we review other studies that have analyzed the interplay between resistance profiles and virulence in clinical strains from the more general point of view of resistance as the final outcome (not considering specific mechanisms). In this section, we consider both in vivo models and in vitro studies performed with clinical strains of P. aeruginosa.

Clinical strains

Focusing on the clinical strains, many authors have attempted to clarify the association between different virulence factors and MDR/XDR profiles. Studies are summarized in Table 3.

Table 3 Studies reviewing the interaction between resistance profiles and virulence in clinical strain collections and through animal models
Full size table

As mentioned above, several studies with clinical strains suggest that the T3SS genotype, [5, 69, 137] and the serotypes [42, 138, 139], are key virulence factors in P. aeruginosa. Particularly, the exoU genotype is the major virulence determinant of P. aeruginosa [5, 38, 42, 46, 47, 69, 140].

Different studies have explored the relationship between other virulence determinants and antimicrobial resistance phenotypes reaching different conclusions. Regarding the ability to produce pigments such as pyocyanin some studies found a negative correlation between MDR profile and pyocyanin production [141, 142] while others did not find association between the resistance phenotype and the pyocyanin production [70, 143,144,145]. Other virulence determinants such as motility or alginate production were also analysed, but contradictory results were observed regarding the association of these virulence determinants and the resistance pattern [141, 144,145,146] as it is shown in Table 3.

With specific reference to the activity of virulence factors conventionally described as highly dependent on quorum sensing (elastase, alkaline protease, pyocyanin and biofilm production), in a single center restrospective study was analyzed the relationship between these and antimicrobial susceptibility in respiratory isolates and was found that those that tested negative for virulence factor production were predominantly more resistant to antimicrobials [148].

Finally, numerous studies have analyzed the relationship between biofilm production and resistance with some studies that found a correlation between MDR/XDR resistance profile and a higher ability to biofilm formation [144, 147, 149] and others showing no correlation between the resistance profile and the biofilm formation [143, 146, 150] or even a negative correlation [151].

These conflicting results in the resistance profiles of clinical strains one more time highlight the significant complexity in the relationship between resistance and virulence.

Animal models

In vitro investigations of human pathogens do not always replicate natural conditions, and the expression of virulence factors is not the same when cultured in vitro as in animals or humans. As a result, animal models have become a valuable tool for advancing our knowledge of P. aeruginosa pathogenesis and host–pathogen interactions as well as for the development and evaluation of new therapies [152, 153]. Although some studies involving animal models have been already mentioned in previous paragraph, in this section we bring together the most important studies dealing with virulence-resistance interplay in P. aeruginosa, in which animal models play a key role.

With respect to vertebrate models, Giamarellos-Bourboulis et al. have used experimental models observing higher mortality, stimulation of innate immunity and bacterial load in infections caused by susceptible strains than by MDR strains [154,155,156].

The authors suggested that their findings may support the idea that some of the differences in virulence between MDR and susceptible P. aeruginosa can be attributed to interactions with the innate immune system. Similar results were found by Gómez-Zorrilla et al. [64, 157].

In general, susceptible P. aeruginosa strains induced a higher inflammatory response than MDR strains, but even strains with a similar resistance phenotype induced different responses. Although their data suggested that there was a fitness cost associated with multidrug resistance, the magnitude of the inflammatory response was the result not only of the resistance phenotype but also of the virulence phenotype. The expression of virulence factors not only affects the virulence of P. aeruginosa but also alters the magnitude of the host–pathogen interaction, accounting for the differences in the elicited inflammatory response [157].

Recently, a study has evaluated the usefulness of a machine learning approach to predict P. aeruginosa virulence in a mouse model of bloodstream infection based on genomic content, using a training set of P. aeruginosa bacteremia isolates. Authors showed that the core genome, alone or in combination with the accessory genome, is predictive of virulence. Thus, the machine learning analyses performed in this study could serve as a framework for further investigation of the relationship between bacterial genomes and their phenotype [159].

As an alternative to the vertebrate models, studies are increasingly being conducted in invertebrate models such as the fruit fly Drosophila melanogaster, the nematode Caenorhabditis elegans, and the wax moth Galleria mellonella [160, 161]. C. elegans studies are a promising model for studying developmental biology and host–pathogen interactions [162,163,164,165]. In this regard, Recio et al. found that the virulence of the P. aeruginosa strains they studied correlated significantly with the non-MDR and MDR phenotype, exoU+ and exoS+ genotypes, O4 and O11 serotypes and the ST175 high-risk clone [69]. Sánchez-Diener et al. documented a clear negative correlation between antimicrobial resistance and virulence by testing a large collection of well-characterized P. aeruginosa isolates from different sources (bloodstream infections, nosocomial outbreaks, cystic fibrosis and environmental samples) in a C. elegans infection model. The lowest virulence was linked to XDR profiles, but virulence varied considerably depending on the high-risk clone involved, being high for ST111 and ST235, but very low for ST175. The highest virulence of ST235 could well be attributed to its exoU + T3SS genotype, linked to higher virulence in the C. elegans model, as well as in previous clinical studies [104]. Therefore, they suggested that the global transcriptional regulator AmpR could be involved in modulating bacterial pathogenicity as shown in previous works [105, 166]. The same authors investigated the impact of virulence phenotype on mortality in 593 P. aeruginosa bloodstream isolates recovered from a prospective Spanish multicenter study in order to assess the predictive value of C. elegans lethality as a prognostic marker. Their results indicated that the virulence phenotype of P. aeruginosa (specifically in the C. elegans model) correlated with virulence genotype (T3SS) and resistance profile but was a poor prognostic marker of mortality in bloodstream infections [167]. On the other hand, previous studies by Miyata and collaborators determined that the greater wax moth caterpillar G. mellonella is also a suitable invertebrate model host for studying the role of T3SS in P. aeruginosa pathogenesis [161]. Moreover, a very good correlation in the virulence outputs showed by 32 different P. aeruginosa strains when comparing G. mellonella with murine models was also found in another study, demonstrating the validity of this invertebrate model [168].

Clinical impact of multidrug resistance in Pseudomonas aeruginosa infections

The clinical impact of multidrug resistance in P. aeruginosa is not only determined by bacterium-related factors, as reviewed in the sections above, but also by factors related to antimicrobial therapy, as well as host-dependent factors such as immune response. Against this background, it is difficult to determine the exact direct relationship between multidrug resistance and clinical outcomes. While it is generally assumed that infections caused by MDR microorganisms are associated with worse outcomes [169,170,171,172], these worse outcomes could be associated with treatment and/or host factors, rather than factors directly associated with bacteria.

With respect to the treatment, one of the main consequences of MDR/DTR infections is the challenge of selecting an appropriate empirical antibiotic therapy. Infections caused by MDR/DTR P. aeruginosa are associated with an increased risk of the patient receiving inappropriate empirical therapy and delayed appropriate antimicrobial therapy [173, 174]. It is well demonstrated that delayed appropriate empirical antimicrobial therapy and decreased antibiotic effectiveness are associated with increased mortality [1, 12, 175,176,177]. In addition, antimicrobial therapies used for MDR/DTR infections may be less effective than those used to treat susceptible infections, often requiring second-line antimicrobial agents or suboptimal treatments [178].

With respect to the host, MDR/ DTR-P. aeruginosa infections usually occur in severely ill patients with multiple underlying diseases, frequently associated with longer hospitalization or previous ICU admissions. These factors may explain, at least in part, the worse outcome of MDR/DTR-infections [179,180,181,182]. In one of the largest studies evaluating the impact of carbapenem resistance on 30-day mortality, Peña et al. demonstrated that carbapenem resistance was associated with significantly increased 30-day mortality rates but that the relationship depended on the degree of baseline severity [183]. Thus, patients with infections due to multidrug-resistant microorganisms tend to have more comorbidities, which may lead to higher mortality [12].

Although a number of studies over the years have compared the different outcomes for infections caused by susceptible and multidrug-resistant strains, the question of whether infections caused by multidrug-resistant microorganisms lead to worse outcomes than those caused by susceptible ones is still being debated. Table 4 lists and summarizes the main previous studies conducted to evaluate differences in clinical outcome between susceptible and drug-resistant P. aeruginosa infections according to source of infection [bacteremia, pneumonia, and urinary tract infection (UTI)], there are conflicting results with some studies suggesting that MDR infections are associated with increased mortality and others showing the opposite.

Table 4 Clinical studies providing outcome information for infections due to MDR/DTR Pseudomonas aeruginosa (first studies supporting association of resistance with mortality and second studies against the association)
Full size table

Apart from relevant clinical variables and inappropriate empirical therapy, different studies have identified an association between antimicrobial resistance and worse clinical prognosis in bloodstream infections [6, 42, 183,184,185,186,187,188,189,190,191,192,193]. In other studies, however the MDR/DTR phenotype was not associated with poor prognosis in bloodstream infections [183, 196,197,198,199].

With respect to respiratory tract infections, most studies have analyzed outcomes in ventilator-associated pneumonia (VAP). Although some reported correlations between different resistant microorganisms other than P. aeruginosa and mortality [200,201,202,203,204], other studies observed worse clinical outcomes associated with MDR pathogens, such as longer intensive care unit (ICU) stay, prolonged duration of mechanical ventilation (MV) and higher rates of microbial persistence, but no impact on mortality [205,206,207]. Concerning MDR/DTR P. aeruginosa infections, associations with longer duration of MV, longer ICU stay and longer delay to appropriate antimicrobial therapy administration have also been noted, but with no significant differences in treatment failure and/or death [208,209,210,211,212,213].

In terms of UTI, few studies have analyzed infection outcomes according to P. aeruginosa resistance patterns. In two retrospective studies it was found that the MDR phenotype was not associated with in-hospital and all-cause late mortality [215, 216].

Regarding the association of virulence determinants and resistance profiles, not many studies have investigated its impact on clinical outcomes. In a multicenter study previously mentioned by Peña et al., including 593 P. aeruginosa bloodstream infections, was investigated the interplay between the T3SS genotypes (exoS, exoT, exoU and exoY genes), the MDR phenotypes, and the influence in mortality. It was observed that exoS + isolates were more frequently MDR than exoU + isolates and exoU genotype was positively related to the moderately resistant phenotype and negatively linked to the XDR phenotype. They found that early mortality was associated with exoU + isolates and late mortality was not conditioned by the T3SS genotype but was independently associated with the MDR phenotype [5]. These results support that T3SS genotype is the most important virulence determinant in P. aeruginosa and exoU + is associated with worse clinical otucomes. Jeong et al. also analyzed, in a study conducted in Korea, the impact of virulence factors on clinical outcomes in 63 CR bacteremia strains. They found that the capacity of the strain to form biofilm was an independent predictive factor for mortality. In addition, the biofilm-forming ability and elastase activity of strains were correlated with the severity of the infection (determined by the Acute Physiology and Chronic Health Evaluation II (APACHE II) score) [217].

Economic burden of multidrug resistance in Pseudomonas aeruginosa infections

Apart from their clinical impact and threat to patient health, P. aeruginosa infections and the interplay between resistance and virulence are associated with significant burden and cost, which are worth reviewing here. The additional cost of a single case of P. aeruginosa infection for example has been estimated to be approximately 18–19,000 Euros [218, 219].

Antimicrobial-resistant infections have been shown to be associated with a higher economic burden compared to those caused by susceptible strains. In a recent meta-analysis and systematic review of the economic cost of resistance (P. aeruginosa was among the three most studied bacteria), the attributable cost of resistant infection ranged from − US$2,371.4 to + US$29,289.1 per patient episode, the mean excess length of stay was 7.4 days, and the odds ratio of readmission for patients with resistant infection was 49.2% higher than for those with susceptible strains [220].

For patients with MDR P. aeruginosa nosocomial infections, the total mean economic cost per admission is higher than for those with non-resistant strains (15,265 vs. 4933 Euros) [221]. For respiratory infections, in a U.S. study, the excess cost per case was US$22,370, the average adjusted excess length of stay was 6.7 days and the adjusted readmission rate was 16.2% versus 11.1% in non MDR infections [222]. Carbapenem-resistant compared to carbapenem-susceptible P. aeruginosa infections are also associated with longer median length of stay, higher total hospital costs (median $6082.0 vs $4954.2) and daily hospital costs (median $236.1 vs $223.6) [223, 224]. Meanwhile, a recent retrospective multicenter study in China analyzing the clinical and economic burden of CR infection or colonization caused by Klebsiella pneumoniae, P. aeruginosa and Acinetobacter baumannii found that the CR phenotype was related to increased total hospital cost in all microorganisms ($4605 in CR P. aeruginosa) and excess length of stay (5.4 days in CR P. aeruginosa) [225].

Finally, it should be also noted the consequences associated with the high cost of the newly antimicrobials introduced in the market for difficult to treat infections or those expected to be approved in the future. In a recent study estimating the annual incremental costs of treating, among others, difficult-to-treat Gram-negative bacteria with new drugs in US hospitalized patients, was estimated that the cost of a 14-day course for MDR P.aeruginosa was doubled with new drugs; and the annual incremental costs of treating difficult-to-treat Gram-negative bacteria ranged from 30 million to over 500 million US$ [226]. This situation is especially worrisome in lower-middle and low-income countries where the clinical and economic impact of antimicrobial resistance is more pronounced [227, 228] and the access to healthcare facilities and new antimicrobials is limited [229].

Therefore, although the question of the clinical consequences of resistant P. aeruginosa infections, particularly the impact on mortality, remains open (with many different studies showing conflicting and controversial results, as shown above), the evidence that antibiotic-resistant infections have a much greater economic impact on healthcare systems is overwhelming. Further research is needed therefore to optimize antimicrobial use and control strategies for multidrug-resistant infections.

Conclusions

Although it has been shown that MDR/DTR P. aeruginosa is associated with an increased economic burden, attributable to prolonged length of stay and higher average total costs, the topic of the biological implications of antibiotic resistance for virulence and fitness in P. aeruginosa, and thus for patient outcome, is currently debated. As we explained above, although it is generally assumed that acquisition of resistant determinants is associated with a fitness cost, several studies support that resistance mutations may not be associated with a decrease in virulence and/or that certain compensatory mutations may allow MDR/DTR strains to recover their initial fitness. Nonethless, mutations also can have environment effects by increasing or decreasing fitness under certain conditions, for example influenced by the antimicrobial treament or the host. Therefore, environmental variability may modify the tendency to compensatory mutations if the fitness of resistance mutations differs among settings.

For the clinical consequences of MDR/DTR infections, it is difficult to establish a direct association between MDR/DTR profile and mortality attributable to the infection. Conflicting results for the clinical outcomes of MDR/DTR P. aeruginosa infections have been reported, with different possible explanations for the inconsistent results, including treatment and host-dependent factors. Patients with MDR/DTR infections are at increased risk of receiving inappropriate empirical antibiotic therapy and alternative therapies (second-line treatments) that are less effective than those used for treating non-MDR infections. In addition, host-dependent factors associated with MDR/DTR infections may also partly explain the worse outcome in these patients, as these infections usually occur in patients with severe preexisting comorbidities, more severe illnesses and longer prior hospital stays. Certain methodological limitations, such as the retrospective study design or studies with small sample sizes may also act as biases. Consequently, further studies are needed to clarify the real effect of infections caused by MDR and DTR P. aeruginosa.

In terms of the basic knowledge reviewed here, a number of conclusions can be drawn. The first is that the interplay between resistance and virulence is highly complex, and is reflected in the remarkable occurrence of results that contradict each other for specific resistance mechanisms. Complexity is also reflected in the variability of costs depending on the mechanism: high burden vs innocuous vs enhancement of virulence. Second, from the information collected, it can also be gathered that, although a lot of data is available, there are several gaps in knowledge. To give some examples, the field of compensatory evolution associated with different resistance mechanisms appears not to be well studied in P. aeruginosa. Another gap that has not been filled is determination of the specific burden associated with most horizontally acquired β-lactamases, as well as of the mechanisms determining resistance to newly introduced drugs (ceftolozane/tazobactam, imipenem/relebactam, ceftazidime/avibactam and cefiderocol) or those likely to be approved in the near future (aztreonam/avibactam, cefepime/zidebactam or cefepime/taniborbactam) [230]. A holistic approach is required to understand all aspects of the balance between resistance and fitness/virulence in P. aeruginosa, including basic research and clinical perspectives. Deep knowledge of the topic is essential to find anti-virulence targets, and to understand the intricate landscape of the co-evolution of resistance and virulence in the context of acute infection. This could help to understand how individual antibiotic resistance mechanisms fine-tune P. aeruginosa virulence-related behavior thereby making it possible to predict modulation in the pathogenesis of infection and the consequences for the patient along over time, knowledge that can be exploited to good effect in clinical practice [231]. It is anticipated that opening up or digging deeper into the fields of research mentioned above will be of great help in combating one of the greatest challenges to public health in the twenty-first century, namely, P. aeruginosa resistance.