Host Stress drives tolerance and persistence; the bane of anti-microbial therapeutics

Abstract

Antibiotic resistance, typically associated with genetic changes within a bacterial population, is a frequent contributor to antibiotic treatment failures. Antibiotic persistence and tolerance, which we collectively term recalcitrance, represent transient phenotypic changes in the bacterial population that prolong survival in the presence of typically lethal concentrations of antibiotics. Antibiotic recalcitrance is challenging to detect and investigate; traditionally studied under in vitro conditions, our understanding during infection and its contribution to antibiotic failure is limited. Recently, significant progress has been made in the study of antibiotic recalcitrant populations in pathogenic species, including Mycobacterium tuberculosis, Staphylococcus aureus, Salmonella enterica, and Yersiniae, in the context of the host environment. Despite the diversity of these pathogens and infection models, shared signals and responses promote recalcitrance, and common features and vulnerabilities of persisters and tolerant bacteria have emerged. These will be discussed here, along with progress towards developing therapeutic interventions to better treat recalcitrant pathogens.

Introduction

Antimicrobial resistance, where pathogenic microorganisms survive chemotherapy, is a significant and expanding challenge to human health. Bacterial pathogens can survive antibiotic exposure through drug resistance or through phenotypic changes that confer reduced drug susceptibility, through persistence or tolerance mechanisms¹ (Fig 1). Both persistence and tolerance, which we refer to collectively as “recalcitrance”, are implicated in treatment failure, but are less well-studied than resistance, due to a paucity of appropriate tools and good experimental models. Recalcitrant bacteria are transiently growth-arrested cells that can survive killing by several classes of antibiotics at concentrations that should be lethal. The eventual resumption of growth by recalcitrants likely contributes to infection relapses. Therefore, we have a pressing need to understand the physiology and vulnerabilities of these bacteria. Ultimately, this knowledge will inform new effective ways to clear bacterial infections and limit the development of antibiotic resistance.

Figure 1-.

Persisters: a small number of bacteria that survive exposure to lethal concentrations of antibiotic, but upon repetition of exposure the same number of survivors emerge. Phenotypic tolerance: is the transient reduction in antibiotic susceptibility that is usually the product of environmental stress, but can impact large proportions of a bacterial population at any given time. A major difference between antibiotic persistence and tolerance is the penetrance of the recalcitrance in the population. Resistance: arises through mutation, is heritable and would apply to all bacteria in a clonal population.

Since the discovery of the persister phenomenon during the treatment of wound infections in the 1940s, studying antibiotic persistence has proven challenging owing to the transience and reversibility of this physiological state. The establishment of single-cell methods that enable the tracking, collection, and analysis of this recalcitrant subset has led to important breakthroughs in recent years, such as the realization that recalcitrant bacteria promote the emergence of antibiotic resistance^2,3. For a long time, there has been a lack of a cohesive understanding of the molecular determinants that underpin the phenomenon of recalcitrance. This is partly because most studies on persisters have been carried out on bacteria grown in isolation in artificial laboratory media. This is limiting for several reasons. First, antibiotic recalcitrance occurs at an excessively low frequency when bacterial populations are not stressed, leaving the scientists vulnerable to the inflated importance of small variabilities in their experimental setups. Second, many different physiological changes have been convincingly shown to enable bacteria to survive antibiotics in vitro making it hard to pinpoint if there is a main path to persistence that is more prevalent than others. This has led to the perception that the phenomenon is purely stochastic and is to remain elusive.

Finally, experiments in bacterial growth media lack the environmental complexities encountered by the pathogen within a host and yield a limited understanding of recalcitrant microbes and their vulnerabilities during infection. Indeed, recalcitrance during infection is more demanding in that it involves more than the ability of bacteria to survive antibiotics, but to weather and respond to the combined assault of antimicrobials and host immune defenses⁴. Hence, although some progress in understanding persistence has been made under in vitro conditions, investigating antibiotic recalcitrance of pathogens during infection is, therefore, more likely to reveal the more relevant paths to persistence amongst all the proposed models. Bacteria typically interact with immune cells early during the infectious process. Respiratory pathogens, such as Mycobacterium tuberculosis, likely encounter alveolar macrophages very early after inhalation. Mucosal pathogens, such as Yersiniae or Salmonellae may initially interact with epithelial cells, which sense the presence of pathogen-associated molecular patterns (PAMPs) with pattern recognition receptors, initiating production of proinflammatory signals that promote the recruitment of immune cells to the site of infection. For pathogens that have spread systemically, such as Staphylococcus aureus, Yersinia or Salmonella, bacteria encounter immune cells both in the bloodstream and within host tissue sites. Bacteria are typically filtered from the bloodstream into the spleen and may simultaneously seed distinct sites such as the liver or kidneys. In both the spleen and liver, bacteria encounter resident immune cell subsets immediately, whereas, in organs like the kidney, interactions with stromal cells trigger subsequent immune cell recruitment. Immune cell recruitment is rapid, and neutrophils arrive at sites of infection to contain bacterial replication within hours of sensing bacterial presence. These immediate or very early interactions with phagocyte populations are meant to contain bacterial proliferation, but successful pathogens use an arsenal of factors to disarm the immune response and promote bacterial colonization of host sites.

Bacteria have distinct interactions with host immune cells depending on whether they proliferate extracellularly or establish an intracellular niche. For example, bacteria that primarily proliferate extracellularly rely on effectors injected by secretion systems, cell surface changes such as polysaccharide capsule production, aggregate or biofilm formation or even an increase in bacterial cell size to prevent phagocytic uptake and establish an extracellular niche. In contrast, intracellular bacteria orchestrate specific changes within a host cell to establish their niche. These changes can also be promoted by secretion systems and their associated effector proteins. Bacteria residing within immune cells can further be contained within an intracellular compartment or reside within the host cell cytosol. No matter the niche, bacteria have to overcome host-derived stressors to survive.

Some major examples of stressors, relevant to many bacterial pathogens, include nutrient limitation, low pH, and reactive oxygen and reactive nitrogen species (ROS, RNS). These are thought to be early host responses to infection, largely produced by innate phagocyte populations. Many pathogenic species, including Mycobacterium tuberculosis, Staphylococcus aureus, Salmonella enterica, and Yersiniae form recalcitrant cells at a high rate during infections because bacteria alter their phenotypes by arresting growth in response to these host stressors^5–10. Several labs, despite working on very different pathogens and disparate infection models, have made significant progress in recent years and shared signals and responses that promote antibiotic recalcitrance, as well as common features and vulnerabilities of persisters and tolerant bacteria have emerged. The phenomenon of recalcitrance is no longer intangible and progress towards interventions now seems within reach. This will be the focus of this review.

Four pathogens

M. tuberculosis, S. aureus, S. enterica, and Y. pseudotuberculosis are very diverse pathogens with different infection routes, niches, and patterns of AMR in patients. They nonetheless face common stressors and share increased antibiotic recalcitrance during infection.

Infection with M. tuberculosis is initiated when the bacterium is inhaled and subsequently phagocytosed by the resident alveolar macrophages that patrol the airway surfaces. The bacilli exploit secreted effector proteins and the host’s response to released cell wall lipids to modulate the host cell and tissue environment to support their survival and growth^11,12. In tuberculosis, multi-drug resistant (MDR) and totally-drug resistant strains (XDR), that are virtually untreatable, are a problem of growing magnitude. The emergence of drug resistance in most bacterial species is problematic primarily because of horizontal transfer of the resistance element across bacterial isolates and species, which is not a problem with tuberculosis. M. tuberculosis lives a relatively solitary lifestyle and most horizontal gene transfer events are ancient and shared across M. tuberculosis species, and while mycobacterial phages are around, and have been exploited as tools for genetic manipulation, there is little evidence of modern genetic exchange in M. tuberculosis. The problem with tuberculosis is that acquisition of drug resistance occurs as independent events with high frequency across M. tuberculosis strains, and as such heritable drug resistance expands linearly. This indicates that, under current treatment regimens, the bacterium is predisposed to develop heritable drug resistance. Effective treatment against active tuberculosis involves treatment with 4 antibiotics, usually isoniazid (INH), Rifampicin (RIF), Pyrazinimide (PZA) and Ethambutol (ETH) for periods up to and exceeding 9 months. This requirement for prolonged multidrug treatment is the product of several properties inherent to M. tuberculosis that we believe have been selected for through evolution. M. tuberculosis evolved from saprophytic actinomycetes^13,14 and was likely to have emerged as a human pathogen around the time human species spread out of Africa 50-70,000 years ago^15,16. M. tuberculosis has no other host of significance therefore its recent evolution has been shaped exclusively through its interaction with humans. However, M. tuberculosis’ relationship with antibiotics likely predates its invasion of humans as a host environment as its ancestors co-habited with soil bacteria that use antibiotics for interbacterial warfare.

S. aureus is a versatile pathogen that can cause infections ranging from superficial skin and soft tissue infections to invasive infections such as endocarditis and bacteremia. Antibiotic treatment of S. aureus infection is frequently complicated with treatment for S. aureus bacteremia requiring 4 to 6 weeks of i.v. antibiotics¹⁷. Despite prolonged therapy, treatment failure occurs in about 1 in 4 patients, leading to around 20,000 deaths in the US each year¹⁸. S. aureus virulence is driven by a multitude of virulence factors but the production of numerous toxins is the key factor in pathogenicity. Alpha-toxin is the best-studied S. aureus toxin. It is a beta-barrel toxin that causes pore formation in eukaryotic membranes. It is particularly important for bacterial penetration through tissue and killing and incapacitation of innate immune cells. Leukocidins are another family of toxins central to S. aureus virulence. These bi-component pore-forming toxins destroy immune cells including phagocytes, dendritic cells, T lymphocytes and natural killer cells^19,20. S. aureus, although classically understood to be an extracellular pathogen, is now considered a facultative intracellular pathogen, capable of surviving and proliferating within a range of host cell types. Neutrophils and macrophages have been particularly well established as significant reservoirs of S. aureus in murine bacteremia models^21,22. Antibiotic resistance has been a major complicating factor in the treatment of S. aureus infection. Most notoriously, methicillin-resistant S. aureus (MRSA) are increasingly common, and they frequently harbor resistance to numerous antibiotics. In recent decades the emergence and spread of community-acquired MRSA (CA-MRSA) of the USA300 lineage has been particularly problematic as these strains exhibit multidrug rehsistance and increased virulence, resulting in the establishment of infection in otherwise healthy individuals. Despite the importance of resistance, most S. aureus isolates remain susceptible to numerous clinically-approved antibiotics, including vancomycin, daptomycin, linezolid and cephalosporins. Despite susceptibility to these antibiotics, treatment failure is common. Importantly, mortality rates associated with methicillin-susceptible S. aureus bloodstream infection and MRSA bloodstream infection are similar¹⁸, so antibiotic failure is clearly complicated and cannot be explained solely by antibiotic resistance.

Three Yersinia species are pathogenic to humans: Yersinia pestis, Yersinia enterocolitica, and Yersinia pseudotuberculosis. Y. pestis and Y. enterocolitica emerged from an ancestral Y. pseudotuberculosis, and like Y. enterocolitica, Y. pseudotuberculosis is an enteric pathogen with broad animal host range^23–26. Y. pestis is also isolated from a wide range of mammalian hosts and their associated fleas²⁷. Enteric Yersinia (Y. enterocolitica and Y. pseudotuberculosis, hereafter Yersinia) are ingested with contaminated food, and initially colonize small intestinal tissues, mesenteric lymph nodes, and Peyer’s patches²⁸. Following bacterial proliferation and tissue damage at intestinal sites, bacteria from the intestinal lumen access the bloodstream and spread to deep tissue sites, such as the spleen and liver²⁹. At both intestinal and deep tissue sites, Yersinia proliferates to form extracellular clusters of bacteria, called microcolonies or pyogranulomas^30–32. Neutrophils are recruited early to the site of infection and circumscribe small centers of bacterial growth. However, Yersinia remains extracellular due in large part to a critical virulence factor, its type-III secretion system (T3SS) and associated effector proteins^33,34. The Yersinia T3SS injects 5-6 distinct effector proteins, termed Yersinia outer proteins (Yops), and collectively these inhibit host cell phagocytosis and cytokine production through uncoupling of the actin cytoskeleton and direct interaction with proinflammatory signaling cascades³⁴. Bacterial T3SS expression and activity are heightened in response to neutrophil contact^30,35, and intoxication by Yops is sufficient to maintain Yersinia in an extracellular niche^28,33. Bacterial clusters proliferate in the presence of neutrophils. Additional host sensing and production of proinflammatory components lead to subsequent recruitment of monocytes, which then confine the layer of neutrophils. Yersinia will continue to proliferate following monocyte recruitment, and if left untreated, patients succumb to systemic infection with an estimated mortality rate of 75-100%^36,37. Progression to systemic infection is relatively rare amongst patients and has been associated with additional underlying disorders, such as hemochromatosis or iron overload disorders, diabetes, and hepatic cirrhosis, in addition to HLA-B27 haplotypes, which has also been associated with chronic inflammatory complications^36,38,39. Antibiotic resistance has been considered relatively rare for Yersinia species. However, the first MDR isolate of Y. pestis, with resistance to all antimicrobials recommended for treatment, was described in 1995, and MDR patient isolates of Y. pseudotuberculosis were characterized in 2017^40–42. For both species, MDR has been linked to the acquisition of plasmids containing multiple resistance genes, and the source of these plasmids is thought to be other Gram-negative Enterobacteriaceae^40,42.

Salmonella enterica enterica is a facultative intracellular pathogen that causes a range of diseases in mammals. The subspecies is further divided into typhoidal (S. Typhi, S. Paratyphi) and non-typhoidal (NTS, S. Typhimiurium, and S. Enteritidis) serovars. Typhoidal strains cause enteric fever, a systemic infection, whereas NTS are responsible for approximately 74 million cases yearly of self-limiting diarrheal infections (dNTS). NTS can however cause invasive non-typhoidal disease (iNTS) in immunocompromised patients and young children with more than 3 million cases per year. Multi-drug resistance is widespread to cephalosporins and fluoroquinolones⁴³. Salmonella is transmitted through the fecal-oral route and after ingestion, invades the intestinal epithelium where it causes local inflammation (dNTS), or crosses the epithelial barrier and reaches systemic sites such as the mesenteric lymph nodes, spleen, and liver (iNTS). Macrophages are a natural niche for Salmonella, where bacteria persist for extended periods in the infected host⁴⁴. Recurrence of iNTS is reported in 20-43% of cases with the same isolate driving relapse in 80% of patients⁴⁵. This situation may arise from persistent sites not cleared by insufficient antibiotic treatment or from restricted reach of the drug, but it could also result from the reactivation of intracellular persister Salmonella. The main virulence factors of Salmonella include a capsule and the typhoid toxin for S. Typhi⁴⁶, and two T3SS called SPI-1 and SPI-2. SPI-1 is mostly used by extracellular Salmonella to inject effectors promoting its internalization by non-phagocytic cells⁴⁷. Once internalized, Salmonella is contained in a vacuole from where it secretes through SPI-2 another set of effectors that target intracellular trafficking and signaling pathways to acquire nutrients and derail the host immune response and metabolism⁴⁸.

Interaction between bacterial pathogens and host immune cells triggers antibiotic recalcitrance

Despite the striking differences between the pathogenesis of the four pathogens, all respond similarly to challenges by host immune cells with a dramatic increase in recalcitrance to antibiotics, whether persistence or tolerance (Fig 2). For reasons discussed in this review and many others, it is generally accepted that antibiotics are not as effective in vivo as they are in vitro. Interestingly, most of the drug optimization efforts, particularly in the field of tuberculosis, focus on improving drug pharmacokinetics and addressing questions of drug penetrance into different infection sites. While these issues are significant, relatively little attention is given to drug recalcitrance, which may represent a greater factor in the frequency of acquisition of resistance. This question has been addressed directly in experiments for all four pathogens.

Figure 2.

In macrophages M. tuberculosis and Salmonella form persisters and tolerant bacteria; S. aureus forms tolerant populations; and Yersinia pseudotuberculosis exposed to macrophages form persisters.

Whereas treatment of S. Typhimurium or S. Enteritidis with bactericidal antibiotics in rich laboratory medium yields 10⁻⁶ persister cells, interaction with host macrophages stimulates persister formation dramatically^5,49. A brief internalization of Salmonella for 30 minutes in murine bone marrow- or human blood monocyte-derived macrophages was sufficient to increase the size of the persister fraction by 1,000-fold. The multi-drug persister population was measured by treatment with cefotaxime, ciprofloxacin, or gentamicin (or a combination of) in a rich medium of the population released from macrophages. Similarly, even brief interaction with macrophages increased the persister population of Burkholderia pseudomallei⁸ or S. aureus⁵⁰.

Sustained interactions with macrophages and concomitant treatments with antibiotics that can reach the intracellular niches have also revealed the dramatic induction of a multidrug-recalcitrant population for S. aureus^9,50, M. tuberculosis⁷ or S. Typhimurium⁵. Tracking the growth status of the persister populations of Salmonella, M. tuberculosis or S. aureus using fluorescent growth reporters^5,7,50–53 showed for all pathogens sizeable fractions of the intracellular populations remaining stuck in growth arrest. It is these growth-arrested populations that contain persisters, but not all growth-arrested bacteria survive leading to successful persistence.

Part of the intracellular populations of Salmonella or M. tuberculosis can successfully proliferate within macrophages. Remarkably, intracellular killing kinetics with cefotaxime, a beta-lactam that readily reaches the Salmonella-containing vacuole, revealed that besides the growth-arrested persister population, the growing population is killed much slower than in rich laboratory medium, owing to its much lower proliferation rate. Thus, for Salmonella, which is a professional intracellular pathogen, two recalcitrant fractions co-exist within host cells: the persisters and slow-growing tolerant bacteria. S. aureus is probably less equipped than Salmonella or M. tuberculosis to manipulate the host intracellular environment, resulting in little to no proliferation in host cells. It is nonetheless hardy and displays sustained intramacrophage survival despite a collapse in metabolic activity associated with more homogeneous antibiotic tolerance^9,54 that may be masking an underlying persister population⁵⁰. The Zinkernagel lab has used clinical samples to show that host-mediated stress could induce a transient antibiotic-recalcitrant state⁵⁵.

In an experimental mouse infection with fluorescent, reporter M. tuberculosis strains⁶, lungs from infected mice were homogenized to generate single host cell suspensions and flow-sorted into activated (CD80high) and resting (CD80low) infected macrophages, which were established in culture and exposed overnight to INH or RIF. It was found that those M. tuberculosis in activated macrophages were far less susceptible to both drugs compared to those in resting macrophages. In agreement with these experiments linking CD80 expression levels with induction of bacterial drug tolerance, activation markers, such as the inverse expression levels of CD11c and the bacteria stress reporter, hspX’::GFP, which expresses mCherry constitutively and GFP when bacteria experience host stress, such as NO or hypoxia^56,57, correlated with the relative sensitivity of M. tuberculosis to the frontline antibiotics INH and RIF in overnight, ex vivo drug exposure of flow-sorted macrophages from an in vivo infection⁵⁸. These data all emphasize the significance of host macrophage diversity in determining bacteria susceptibility to anti-TB drugs in vivo, and how immune-mediated antimicrobial activity can work counter to drug action. Clearly, the environments experienced by M. tuberculosis in the different host macrophage populations present in the infected lung play an extremely significant role in in vivo drug efficacy.

Recently a Y. pseudotuberculosis mouse model of systemic infection was adapted to study antibiotic treatment efficacy^10,59. Following the establishment of clusters of infection in intestinal and deep tissue sites, mice reach morbidity endpoints ~48 hours after monocyte recruitment if left untreated¹⁰. It was shown in this infection model that treating with doxycycline in the hours immediately following monocyte recruitment is sufficient to significantly prolong mouse survival. Treated mice experience a 90% reduction in bacterial load in the first 4h after treatment but the number of surviving bacteria remains steady without further reduction despite high levels of antibiotics within tissues¹⁰. When antibiotic concentrations wane, bacterial growth resumes, resulting in infection relapse. Here, host-derived stress is heterogeneously experienced by the bacterial population, and only a subset of the population becomes transiently tolerant to antibiotics amongst a predominantly antibiotic-susceptible population; this is an example of antibiotic persistence¹⁰. Measuring bacterial growth rate with a fluorescent reporter revealed a sharp decrease in bacterial proliferation after 72h of infection^60,61, at which point antibiotic treatment is not sufficient to prolong mouse survival¹⁰.

Recalcitrants are formed in response to common stressors

RNS/ROS

The most thoroughly described mechanism of macrophage-induced antibiotic tolerance in S. aureus involves the production of oxygen and nitrogen free radicals during the respiratory burst (Fig 3). Upon phagocytosis, macrophages produce a variety of reactive oxygen and nitrogen species (ROS and RNS), including superoxide, hydrogen peroxide, nitric oxide and peroxynitrite among others. These molecules can oxidize and nitrify lipids, proteins and DNA, causing extensive damage and frequently inducing bacterial cell death. However, S. aureus is exceptionally well equipped to tolerate ROS/RNS through an impressive repertoire of antioxidants including two superoxide dismutases SodA and SodM, catalase, and antioxidant molecules such as the abundant pigment molecule staphyloxanthin⁶². This antioxidant response contributes to the survival of S. aureus within a mature phagolysosome⁶³. Interestingly, these surviving bacteria demonstrate almost complete tolerance to bactericidal antibiotics.

Figure 3:

Summary of the shared signals and stresses that trigger recalcitrance in macrophages.

While neutrophils appear effectively pacified by the pathogen T3SS, RNS produced by iNOS⁺ monocytes impart significant stress on Yersinia^32,64,65. In the microcolonies that form, monocytes are prevented from direct access to bacteria because the neutrophils circumscribe and confine the bacilli forming a first ring. Monocyte-produced nitric oxide diffuses from the outer iNOS⁺ monocytes to impact inner bacterial clusters³⁰. Bacteria respond by expressing a nitric oxide detoxifying gene, hmp, which has been used as a marker to detect NO stress. hmp expression specifically occurs at the periphery of microcolonies, and Hmp detoxification of NO prevents NO diffusion into the center of microcolonies, allowing most bacterial cells to proliferate in the absence of this stress^30,60. It was recently shown that NO stress is sufficient to inhibit bacterial growth^60,61,66, and this results in preferential survival of NO-stressed Yersinia in a mouse model of doxycycline treatment¹⁰. Similarly, Salmonella persisters, although formed very early on during interaction with macrophages because of a multiplicity of stresses, are then subsequently locked in their state of growth arrest by macrophage production of RNS. Although the RNS-intoxicated persisters survive penicillin to high levels, they are susceptible to ciprofloxacin⁶⁷. Persisters formed in iNOS-deficient (Nos2^−/−) macrophages resume growth much faster than those in WT cells, so fast that they are cleared by antibiotic treatment over 24 h. When the host ceases production of the reactive species, which are also highly toxic to host molecules, Salmonella persisters resume growth in a slow and heterogeneous manner⁶⁸. In the mouse model of M. tuberculosis infection, the use of iNOS-deficient murine strains abrogated the induction of tolerance, pointing to RNS as being a dominant driver of drug recalcitrance⁶.

Although the effects of RNS are probably pleiotropic, we found that the central metabolism of S. aureus surviving the respiratory burst was significantly damaged⁹. Notably, nitration of aconitase caused a collapse in TCA cycle activity, reduced respiration, leading to ATP depletion, and increased antibiotic tolerance of S. aureus sequestered within the phagolysosome^9,54. This nitration of aconitase appears to be predominantly driven by peroxynitrite, a RNS created when NO reacts with superoxide. The link between host produced ROS, ATP depletion and antibiotic tolerance in S. aureus was further supported by independent work by the Van Bembeke lab⁶⁹. Importantly, reduction of ROS/RNS either through mutation of Ncf, a component of the Phox complex or through oral administration of the antioxidant TEMPOL, improved antibiotic killing in a murine model of S. aureus bacteremia⁹ . Analogous observations were made in Salmonella, where the primary effect of RNS on persisters seemed to be the intoxication of their TCA cycle, through the corruption of the Fe-S containing a-ketoglutarate dehydrogenase complex, decreasing cellular respiration and accompanying ATP production. This reduction in ATP production maintains bacteria in growth arrest, which underpins survival to beta-lactams but decreases the activity of efflux pumps that are needed for persisters to survive fluoroquinolones⁶⁷. Administration of an iNOS inhibitor stimulated persister regrowth, improved antibiotic efficacy, and reduced relapse in a murine model of Salmonella systemic infection⁶⁸.

Metabolic perturbations

In more recent analysis of the M. tuberculosis /host macrophage interplay in vivo by Dual RNA-seq and scRNA-seq it was noted that bacteria in activated, or hostile, host macrophages exhibited stress profiles linked to a range of noxious stimuli in addition to NO, including iron deficiency, redox stress, DNA damage, and metabolic stress, due to accumulation of certain metabolic intermediates^58,70,71. The iron response in both bacterium and host was particularly noticeable⁷¹, with marked up-regulation of iron sequestration genes in proinflammatory monocyte-derived macrophages matched by the up-regulation of genes encoding M. tuberculosis siderophores.

This contrasted with the increased expression of iron exporter genes by alveolar macrophages, matched by an up-regulation of the bacterial iron storage gene, bfrB, to manage an iron excess. The significance of iron to M. tuberculosis survival is well documented^72–74 and it exposes potential targets for adjunctive therapy⁷⁵. Other bacterial pathways impacting drug susceptibility link back to major stress regulons identified previously^6,76,77. For example, DosRS/DevRs is a two-component regulator sensor effector kinase identified in response to growth arrest due to hypoxia. The DosR regulon comprises approximately 48 genes, including hspX used to construct the M. tuberculosis reporter strain discussed earlier⁵⁶, that are primarily involved in limiting the rate of bacterial growth under noxious stimuli^78,79. Transcription of DosRS-dependent genes is highly up-regulated upon macrophage invasion but shows reduced expression upon the resumption of growth following acclimatization to the intracellular environment. Treatment of M. tuberculosis with hypoxia, NO, acid stress or nutrient starvation in vitro all leads to the acquisition of a non-proliferative state that is DosRS-dependent, and in this state the bacterium is markedly recalcitrant to high concentrations of anti-TB drugs. Recently, small molecule inhibitors of both DosR and DosS were identified^80,81. Some of these inhibitors decreased M. tuberculosis survival and reduced drug tolerance to INH under hypoxic conditions. They also impacted lipid metabolism through repression of the triacylglycerol synthase, tgs1, suggesting a broader impact on bacterial metabolism. It has long been known that M. tuberculosis favors lipids and fatty acids to fuel bacterial growth in vivo and knockouts in genes of the methylcitrate cycle (MCC), such as isocitrate lyase (icl1), impact bacterial fitness through accumulation of MCC metabolic intermediates^82,83. Quinonez and colleagues have shown recently that the metabolic stress associated with the accumulation of MCC intermediates in prpD- and icl1-deficient mutants also leads to reduced drug susceptibility to both bedaquiline and INH⁸⁴. Similarly, in S. aureus and Salmonella, in addition to free radical-mediated induction of recalcitrance, nutrient availability, and starvation have also been implicated. Activation of the stringent response contributes to persister formation following phagocytosis in both pathogens with mutants lacking the stringent response demonstrating reduced persister numbers^5,50. Interestingly, the stringent response had however no impact on S. aureus persisters in vitro⁸⁵. Recently, we showed that macrophages limit glucose availability to S. aureus in the macrophage cytoplasm, resulting in the maintenance of antibiotic tolerance even after bacterial escape from the intensely stressful environment of the phagolysosome⁵⁴. Interestingly, this sequestering of glucose was due to an inflammasome-mediated activation of glycolysis in the macrophage in response to S. aureus, resulting in reduced S. aureus glucose availability⁵⁴.

Intraphagosomal stresses

Finally, the intraphagosomal environment has been shown to have a direct effect on recalcitrance induction in M. tuberculosis and Salmonella through the impact on bacterial redox balance. While M. tuberculosis- or Salmonella-containing phagosomes exhibit arrested maturation and only partially acidify^86–88, phagosomal acidification affects drug susceptibility⁸⁹. This pH dependency is consistent with the activation of PhoPR and the pH-dependent shift observed upon uptake of M. tuberculosis into macrophages⁹⁰. Kreutzfeldt and colleagues recently conducted a genetic screen to identify genes that modulated intracellular M. tuberculosis’ susceptibility to INH. They found that mutations in cinA conferred increased sensitivity to INH exposure in intracellular bacteria but not in bacteria exposed to drug in rich media⁹¹. They then demonstrated that knockouts in cinA also showed decreased tolerance to Ethionamide, Delaminid (DEL) and Pretomanid (PMD). CinA appears to mediate its activity through cleavage of the adducts formed between NAD and the drugs INH, DEL, and PMD by means of its pyrophosphatase domain.

Thus, many routes and stresses lead to the induction of drug recalcitrance in pathogens within host cells, and taken together, these findings indicate that phagocytes are well-equipped to restrict bacterial metabolic activity which, in turn, limits bacterial proliferation. These stresses however also undermine the bactericidal capacity of conventional antibiotics, the majority of whose activity correlates with bacterial metabolic activity.

How to target them better?

Despite the challenges posed by immune cell induction of antibiotic tolerance in pathogens, our growing understanding of the phenomenon presents opportunities to target and eradicate this persistent antibiotic recalcitrant reservoir.

Although broadly reducing ROS/RNS levels improves the bactericidal activity of antibiotics in vivo, there are major drawbacks to this approach. Firstly, ROS/RNS play an important role in immunity to the establishment of infection. Patients with mutations in genes encoding the Phox complex suffer from chronic granulomatous disease and have a far higher propensity for developing infection with numerous microorganisms, in particular S. aureus⁹².

ROS/RNS are also important in host cell-cell signaling and the arbitrary removal of ROS/RNS using high dose antioxidants may have deleterious consequences outside of the macrophage or innate immune response^93,94 . It may be possible to target phagocyte-produced reactive species to remove it specifically from the compartment or within the time window where we need to improve antibiotic efficacy. For instance, activating the macrophage antioxidant response may remove the intracellular ROS and facilitate improved antibiotic-mediated killing of intracellular S. aureus. In support of this, nanoparticles containing Nrf2 activators could be efficiently phagocytosed, rapidly reducing intracellular ROS/RNS and sensitizing intracellular S. aureus to antibiotic killing, in cultured macrophages⁹⁵. Additionally, blocking NLRP3 inflammasome activation with the small molecule, MCC950, resulted in superior rifampicin killing of S. aureus in a murine bacteremia infection model⁵⁴ . As research on the in vivo determinants of antibiotic tolerance continues, additional immunomodulation strategies to sensitize intracellular pathogens to antibiotic killing will likely come to light. Eventually, the identification of an immunomodulator strategy that both improves immune clearance and antibiotic susceptibility in the pathogen may represent a significant breakthrough in the treatment of infections.

In addition to the modulation of the immune response, direct targeting of low-energy, antibiotic-tolerant cells may be promising. Acyldepsipeptides (ADEPs) kill persister cells and eradicate S. aureus biofilm in vitro and in vivo⁹⁶ . The capacity of ADEPs to penetrate host cells and to kill efficiently in the intracellular environment remains to be determined. Additionally, membrane-acting agents that destabilize the cell envelope have been shown to have activity against persister populations of S. aureus in growth medium. Daptomycin is effective at killing S. aureus in stationary phase and it can kill 99% of S. aureus in monocyte-derived macrophages in a mouse model of chronic infection. However, growth arrest can lead to the survival of a tolerant sub-population due to cell-wall remodeling⁹⁷ and the efficacy of daptomycin in removing the intracellular reservoir in vivo remains to be determined. Development of new small molecules that kill non-proliferative, low-energy S. aureus will likely yield antibiotics with improved efficacy not only against immune cell-induced antibiotic-tolerant bacteria, but also the highly tolerant bacteria within biofilms, which also contribute to recalcitrant S. aureus infection associated with implanted devices. The identification of key vulnerabilities of recalcitrant bacteria helps design tailored interventions. The strongest discriminating feature of Salmonella persisters during infection is the DNA damage they experience, particularly double-strand breaks (DSB), which may partially explain their growth arrest. Repairing DNA damage is thus key for persisters to resume growth and lead to infection relapse, and mutants of Salmonella unable to carry out repair of DSB by homologous recombination are highly defective in antibiotic recalcitrance and infection relapse⁹⁸. The improvement of current inhibitors of RecA-driven homologous recombination may lead to a potent adjunctive to antibiotics.

Host-directed therapeutics (HDTs) is an expanding area of interest, particularly in tuberculosis research⁹⁹. The majority of published studies in this sphere have focused on the activity of host-active compounds in monotherapy treatment regimens in culture or in murine models. Most researchers have adopted a candidate-based approach targeting known anti-M. tuberculosis pathways linked to increased anti-microbial activities^100,101, iron and cholesterol distribution^102,103, enhanced induction of autophagy^104,105 and cytokine agonists or antagonists⁹⁹. While several studies do report anti-microbial efficacy, the impact of most HDTs is generally relatively modest. Moreover, earlier studies indicate that host pathways that induce bacterial stress may also lead to the induction of drug tolerance in M. tuberculosis⁶. However, a few studies have explored HDTs in the context of anti-TB drug therapy. Some approaches address modulation of host metabolism^102,103, others enhanced drug accessibility, such as metalloproteinase inhibitors that stabilize the vasculature in the granuloma¹⁰⁶, and others are exploring the impact of the Type 2 diabetes drug Metformin that activates AMPK, and has more pleotropic effects on host cell programming^107,108. There are currently adjunctive human therapy studies ongoing with Metformin in combination with existing anti-TB drug regimens^{107,109–112} that promise to be extremely informative as to the feasibility of combinatorial treatment with HDTs together with frontline anti-TB drugs.

Finally, it is interesting to consider how metabolic stimulation of antibiotic recalcitrant bacteria may result in the sensitization to antibiotics. For instance, the delivery of an appropriate sugar that can be metabolized via glycolysis, to push a non-respiring, antibiotic-recalcitrant populations into a higher energy, antibiotic-sensitive state. Increasing glucose availability to intracellular S. aureus can sensitize them to antibiotics, even under oxidative stress⁹. Interestingly, mannitol was previously administered with aminoglycosides to improve antibiotic killing of E. coli in mice, and fructose increased aminoglycoside susceptibility in S. aureus in culture¹¹³. Direct delivery of fructose and most other sugars to “wake-up” and sensitize persisters is complicated by the host cell capacity to compete for and consume the same sugars.

Future aspects/questions

Despite major advances in our understanding of antibiotic recalcitrance in pathogens during infection over the past decade, many major questions remain.

Although the determinants of antibiotic recalcitrance of very different pathogens overlap as we reviewed here, these pathogens were all considered in a similar infection microenvironment: the host immune cells. The relative contribution of antibiotic-recalcitrant populations associated with innate immune cells, embedded in a biofilm, or within an abscess remains to be investigated and underlying determinants will very likely differ in distinct niches of recalcitrance (Fig 4).

Figure 4:

Greatest challenges for chemotherapy for each infection. M. tuberculosis: differential drug susceptibility driven by heterogeneity in the host environments (granuloma caseum, macrophage subsets). Salmonellae: evolution of resistance from recalcitrance. Yersiniae: strep resistance/MDR and recalcitrance. S. aureus: resistance and biofilms

In tuberculosis infection, the heterogeneity of the host environments undoubtedly impacts the responsiveness of the bacteria to anti-tuberculosis drugs. Targeting non-growing bacilli within the caseous centers of granulomas is linked to the penetrating characteristics of the drug and is being pursued actively by investigators^114–117. But the heterogeneity of the host macrophage populations, whether tissue-resident, alveolar macrophages, or blood monocyte-derived, interstitial macrophages, has received less attention. Adjunctive therapy with host-directed therapeutics to either enhance and/or normalize anti-TB drug activity through reprogramming macrophage metabolism, such as reported for Metformin¹⁰⁸, represent an attractive route forward worthy of additional study.

Additionally, although we have identified common host stressors that can induce antibiotic recalcitrance, we have made comparatively less progress in identifying bacterial pathways that contribute. Screening a panel of E. coli isolates from patients with recurrent bacteremia we identified mutations occurring within patients that led to increased antibiotic recalcitrance in vitro and in vivo. This study was important to demonstrate that recalcitrance can be selected for in patients and that it limits antibiotic efficacy¹¹⁸. With investigations on antibiotic recalcitrance in assays that mimic an infection microenvironment of interest, we predict that new pathways of antibiotic tolerance and persister formation will be identified, that differ markedly from persister formation under artificial, nutrient-rich, and relatively stress-free conditions^98,119.

Although understanding what mediates antibiotic recalcitrance in all of these microenvironments is important, the contribution of persister cells and antibiotic-tolerant populations to poor clinical outcomes has yet to be fully determined. It is likely that recalcitrant phenotypes, antibiotic penetration to certain sites of infection, sub-optimal dosing and ineffective immune control all contribute to treatment failure, and determining their relative importance and possible redundancy, particularly in human disease, will be essential to identify those factors critical to its resolution. A combination of murine infection models and clinical sampling, coupled with histology, mass spectrometry, transcriptomics, and immune phenotyping should help address this problem moving into the future.

Antibiotic resistance is a universally-acknowledged public health problem. Antibiotic recalcitrance reduces bacterial susceptibility to antibiotics and is a major contributor to treatment failure. Here we report an integrated understanding of the signals and responses that promote recalcitrance of the four pathogenic species Mycobacterium tuberculosis, Staphylococcus aureus, Salmonella enterica, and Yersiniae.