Infection therapy: the problem of drug resistance – and possible solutions

Summary The rising antibiotic resistance in major bacterial pathogens together with the breakdown of the antibiotic discovery platform creates a critical situation for infection therapy. Recent developments reviving new antibiotic discovery from defining chemical rules for membrane‐passing compounds to isolation chips for soil bacteria and exploring the human microbiome for antibiotic‐producing bacteria are discussed. The potential of bacteriocins, tailocins, phage lysins, phages, probiotics and commensal blends as alternatives to antibiotics is evaluated.


Breakdown of the antibiotic discovery platform
Then there is an industrial problem with the development of new antibiotics. One aspect is economical: the potentially rapid resistance development against newly marketed antibiotics and their careful, that is limited, use by physicians causes problems with the return of investment for a costly drug development programme. Another aspect is technological, namely the breakdown of the once successful antibiotic discovery 'Waksman platform' (Lewis, 2012(Lewis, , 2013(Lewis, , 2017. By overmining soil-derived actinomycetes, mostly Streptomycetes, for antimicrobial activity, the pipeline run dry and was replaced in the 1960s by a strategy modifying existing compounds into active analogues. The following 'Genomics-Combichem-High Throughput-Rational Drug Design' phase was a disappointment as it did not yield a single new antibiotic drug even after screening 500 000 synthetic compounds (Payne et al., 2007). Antibiotic experts have diagnosed a number of problems. There are only rare chemical compounds that penetrate the lipopolysaccharide-coated outer membrane of Gram-negative bacteria, which limits approaches based on the screening of chemical libraries. In vitro antibacterial tests are another bottleneck as too many pharmacologically important drug characteristics are not measured in this system. Lewis (2017) mentioned other limitations: for practical and commercial reasons, broad-spectrum antibiotics were targeted, while species-selective compounds should be preferred as they avoid collateral damage on the commensal microbiota and cause less selection for widespread resistance development.

Glimmers of hope
Possible ways out of this dilemma include combination therapies, chemical reworking of antibiotics that were discarded for toxic side-effects, or targeting highly connected webs of protein and gene interactions in the bacterial cell that go beyond the few classical antibiotic targets, i.e. ribosomes, penicillin-binding proteins and DNA gyrase/topoisomerase (Brown and Wright, 2016). Furthermore, ninety per cent of natural product chemistry is encoded by silent bacterial operons, whose expression could enable novel discovery platforms (Bentley et al., 2002;Laureti et al., 2011;Rutledge and Challis, 2015). The Gram-negative outer membrane barrier might be crossed by permeable prodrugs that are activated inside the bacterial cell by bacterial enzymes. In addition, chemists have recently defined a set of chemical characteristics needed for the penetration of the outer membrane: the compounds must contain an amine, be amphiphilic and rigid. With these rules, they succeeded to convert an antibiotic active against Gram-positive bacteria for use against E. coli (Richter et al., 2017). Caenorhabditis worms that can be infected with human pathogens by ingestion have been suggested as alternative screening system to in vitro tests which offers several in vivo characteristics, but can still be adapted to microtitre test format (Moy et al., 2006).
Novel antibiotics can also be identified by culture-independent methods as demonstrated by cloning soil microbial DNA libraries (metagenomes) in E. coli (Gillespie et al., 2002). Some of the 'uncultivatable' soil bacteria can be grown inside diffusion chambers incubated in situ, where diffusion provides bacteria with their naturally occurring growth factors (Kaeberlein et al., 2002). Subsequently, these devices were developed into isolation chips allowing high-throughput parallel cultivation platforms (Nichols et al., 2010). When such iChip test systems were inoculated with diluted soil samples and placed back into soil and screened after growth for antimicrobial activity on plates overlaid with S. aureus, a potent new antibiotic was identified: teixobactin. It bound to lipid precursors of cell wall components, causing lysis of many pathogens from Firmicutes and Actinobacteria. Notably, no resistance evolved against this compound possibly because it targets lipids and not proteins and because the producer strain protects itself against the toxic antibiotic by the physical barrier of the outer membrane and not by a genetically encoded antitoxin system (Ling et al., 2015).

. . .to the human microbiome
Interbacterial competition is also a characteristic of the human microbiota, making it a possible source for new antibiotics. Indeed, Staphylococcus lugdunensis, a competitor to S. aureus in the anterior nares of humans, produces a non-ribosomal peptide lugdunin. It inhibits MRSA and vancomycin-resistant Enterococcus isolates with minimal inhibitory concentrations (MIC) in the micromolar range. Lugdunin inhibits DNA, RNA, protein and cell wall synthesis of S. aureus, killing the pathogen without allowing resistance development. Lugdunin is active in an animal infection model (Zipperer et al., 2016).
A systematic analysis of > 2000 reference genomes of the human microbiota identified > 3000 small-molecule biosynthetic gene clusters. The dominant class of small molecules was saccharides, followed by polyketides, non-ribosomal peptides and modified peptides (Donia et al., 2014). To demonstrate the use of this approach, the researchers isolated a thiopeptide from a vaginal Lactobacillus gasseri isolate, which inhibited S. aureus, pathogenic oral streptococci and vaginal pathogens, but notably not vaginal Lactobacillus commensals (Donia et al., 2014).
Microbiologists knew for many years bacteriocins, bacterially produced peptides that are active against other bacteria and against which the producer has specific immunity mechanisms. In the past, bacteriocins were technologically explored for food protection, and nisin has been approved as a food preservative. Subsequently, bacteriocins could facilitate the introduction of probiotic bacteria into an already occupied niche (Dobson et al., 2012). This concept can be extended against antibiotic-resistant gut pathogens: Enterococcus faecalis replaces indigenous, vancomycin-resistant enterococci in mice without perturbing commensals (Kommineni et al., 2015). However, in vitro killing efficacy has not always translated into in vivo protection (Dobson et al., 2012). In vivo efficacy data are still scarce: Lactobacillus salivarius protected mice against invasive Listeria monocytogenes infection (Corr et al., 2007). Thuricin produced by Bacillus thuringiensis kills a wide range of Clostridium difficile isolates in a colon model (Rea et al., 2010) without significant impact on the human gut microbiota (Rea et al., 2011). Bacteriocins are also produced by E. coli (colicins) or Pseudomonas (pyocins), and prevent E. coli diarrhoea in pigs and Pseudomonas aeruginosa infection in wax moths respectively (Behrens et al., 2017). Microcins produced by the probiotic E. coli Nissle strain limited the expansion of competing Enterobacteriaceae in mice suffering from intestinal inflammation and might thus represent a possible treatment mode for enterobacterial colitis (Sassone-Corsi et al., 2016).

. . .to tailocins
Another group of potentially interesting antimicrobials are tail structures from defective phages that function as bacteriocins (tailocins). R-type pyocins resemble the contractile tail of P2-like myoviruses, while F-type pyocins look like the flexible tails from phage lambda. Host specificity is mediated by tail fibres, and when tailocins undergo conformational changes after adsorption to the target bacterium, lethal membrane damage ensues via dissipation of the membrane potential. Efficacy was reported against P. aeruginosa in a mouse peritonitis model (Scholl and Martin, 2008). When the tail spike protein of this pyocin was replaced by a coliphage protein (Scholl et al., 2009), the engineered pyocin showed protection against E. coli infection in a rabbit model (Ritchie et al., 2011). Subsequently, pyocins have been engineered that prevented colonization of mice with C. difficile (Gebhart et al., 2015).
Interesting antimicrobial properties are also displayed by phage lysins (PL; Pastagia et al., 2013). PL attack the bacterial cell wall from the inside to release progeny phage. PL lyse Gram-positive cells also from the outside, while Gram-negative bacteria are protected by the outer membrane. The modular structure of PL with a catalytic and a binding domain connected by a short linker offers mix and match possibilities for bioengineering to increase solubility, thermostability, binding specificity and catalytic efficiency. Most PL show genus-restricted killing activity, sometimes species-restricted killing activity. As they target essential cell wall structures, no PLresistant have yet been described. Only non-neutralizing antibodies develop after in vivo application. PL showed a very rapid and efficient in vitro lytic activity. In vivo activity was described in animal models of pneumonia, endocarditis, pharyngitis, meningitis and sepsis (Pastagia et al., 2013). PL eliminated bacterial pathogens from mucosal and skin surfaces and were active against biofilms. Toxic effects are limited to systemic cytokine induction due to release of bacterial debris. Synergistic activity with antibiotics was observed, and PL could even re-sensitize pathogens to formerly inefficient antibiotics (Daniel et al., 2010). Research is underway to extend their use also against Gram-negative bacteria (Lukacik et al., 2012). Overall, this sounds too good to be true. What are potential drawbacks? PL have a relatively high minimal inhibitory concentration (40 lg ml À1 ) (Daniel et al., 2010) and a very high-affinity constant K a = 10 8 for the cell wall (Loessner et al., 2002). This means that after binding and enzymatic action, PL will not diffuse away to renewed action. A solution might be truncated PL consisting only of the catalytic domain. Also the PL binding domain alone can protect mice from MRSA infection (Raz et al., 2017). Another problem might be that PL are quickly cleared from systemic circulation (half-life in rats: 20 min; Entenza et al., 2005), which would limit its in vivo application, but might be partially compensated by its rapid mode of action.
One might ask why PL in view of these promising activities were not yet tested in clinical trials and whether this represents not a silent argument against their practical value. However, this lack or as in the case of phage therapy, the scarcity of clinical evidence might tell us more about the difficulties to finance product developments with these phage-based approaches than about objective technical hurdles. In view of the public health importance of these potential alternatives in an era of antibiotic resistance crisis, one might ask why phage and PL trials are not organized by governmental agencies.
Phages are bacterial viruses, which infect bacteria, produce progeny phage and lyse their target cells. In theory, we have here a self-amplifying, generally species-specific antibacterial agent in situ produced in proportion to the pathogen level. These are unique pharmacokinetic properties coming close to Paul Ehrlich's magic bullet. Phage therapy has been developed in the Soviet Union as an alternative to antibiotics (Sulakvelidze et al., 2001). Today, a multitude of phage products with a wide range of applications are sold as registered, over-the-counter products in pharmacies of many countries from the former Soviet Union. Scientific reports on the clinical efficacy of phage therapy go back to the pioneers of phage therapy (d'Herelle with cholera; Br€ ussow, 2017) or to a large controlled clinical trial of the Eliava Institute in Georgia in the 1960s, which reported a successful prophylaxis trial against Shigella dysentery and E. coli diarrhoea (Sulakvelidze et al., 2001). Beyond these reports, successful published experience with phage therapy is so far limited to small controlled clinical trials (P. aeruginosa otitis externa infection) or uncontrolled series of phage applications in Polish patients not responding to antibiotics (Vandenheuvel et al., 2015). A controlled clinical trial of E. coli diarrhoea in children from Bangladesh showed no advantage of two different phage preparations, including a commercial Russian product, over standard therapy. Subsequent microbiota analysis revealed that E. coli pathogen titres in the stool of the patients were probably below the replication threshold for the orally applied phages (Sarker et al., 2016). While phage therapy remains an attractive concept, its documentation by clinical trials is still insufficient to confirm its value as an alternative to antibiotics. Phage-bacterium interaction within the human host is largely undefined, and basic questions lack still an answer: What bacterial infections are suitable targets for phage therapy? What phage types present suitable in vivo properties for clinical trials? . . .and bacteriovores The current overview started with small chemical compounds (antibiotics), continued with proteins (bacteriocins, lysins), protein complexes (tailocins) and viruses (bacteriophages) as antimicrobial agents. This gradient of biological complexity for antimirobial agents can be extended to viable bacteria. A particularly fascinating approach is bacteria eating bacteria (bacteriovores). The Bdellovibrio and Micavibrio bacteriovores represent such living antimicrobial agents. Bacteriovores seem to be safe: Rectal bacteriovore application in rats induced only a modest cytokine response and had minimal effects on faecal microbiota composition (Shatzkes et al., 2017b). Zebrafish infected with Shigella flexneri and exposed to Bdellovibrio showed an initial bacterial predation phase and a subsequent elimination phase of the bacteriovore by host neutrophils (Willis et al., 2016). Upon intranasal inoculation in rats, the predator bacteria reduced K. pneumoniae burden in the lungs by 1000-fold (Shatzkes et al., 2016). In contrast, predator bacteria were unable to reduce K. pneumoniae titres in the blood of intravenously infected rats (Shatzkes et al., 2017a). Initial claims for treatment effects against keratitis of the eyes were not confirmed (Boileau et al., 2016), and clinical exploration has thus not yet started.

Outlook
So far, all alternative anti-infectious agents described in the present overview were selected for their lytic action on pathogenic bacteria. Alternative approaches also need to be considered such as measures reducing the colonization level with antibiotic-resistant bacteria like MRSA before it comes to an infection. Ecological interventions reducing the transmission of antibiotic-resistant bacteria belong into this category as well as means to modify the expression of virulence factors by these bacteria. Ecological theory might thus identify novel treatment approaches against polymicrobial diseases with antibiotic-resistant pathogens (Quinn et al., 2016). At the end, I want to mention one such approach: Lactobacillus paracasei strain ST11 showed therapeutic effects in a controlled clinical trial against childhood diarrhoea (Sarker et al., 2005) in a country where diarrhoea-associated bacteria are largely antibiotic-resistant (Jiang et al., 2002). Unfortunately, the mode of probiotic action is only poorly defined (Servin, 2004). Also a clinical trial with a defined cocktail of bacterial gut commensals showed efficacy against C. difficile infection (Tvede and Rask-Madsen, 1989). It is thus likely that we will see more such clinical trials with probiotics or commensal cocktails in infection therapy (Li evin-Le Moal and Servin, 2014). The next level of complexity for antimicrobial agents has already been documented, namely the transfer of an entire microbiota from a healthy subject to C. difficile patients (van Nood et al., 2013). These approaches open a new vista where pathogens are not targeted for lysis, but are put under competitive ecological pressure (Br€ ussow, 2007). As these interactions are complex in nature, it can be hoped that they are less prone to subversion by resistance development of pathogens.