Predicting antibiotic-associated virulence of Pseudomonas aeruginosa using an ex-vivo lung biofilm model

Background Bacterial biofilms are known to have high antibiotic tolerance which directly affects clearance of bacterial infections in people with cystic fibrosis (CF). Current antibiotic susceptibility testing methods are either based on planktonic cells or do not reflect the complexity of biofilms in vivo. Consequently, inaccurate diagnostics affect treatment choice, preventing bacterial clearance and potentially selecting for antibiotic resistance. This leads to prolonged, ineffective treatment. Methods In this study, we use an ex-vivo lung biofilm model to study antibiotic tolerance and virulence of Pseudomonas aeruginosa. Sections of pig bronchiole were dissected, prepared and infected with clinical isolates of P. aeruginosa and incubated in artificial sputum media to form biofilms, as previously described. Then, lung-associated biofilms were challenged with antibiotics, at therapeutically relevant concentrations, before their bacterial load and virulence were quantified and detected, respectively. Results The results demonstrated minimal effect on the bacterial load with therapeutically relevant concentrations of ciprofloxacin and meropenem, with the later causing an increased production of proteases and pyocyanin. A combination of meropenem and tobramycin did not show any additional decrease in bacterial load but demonstrated a slight decrease in total proteases and pyocyanin production. Conclusions We demonstrate a realistic model for understanding antibiotic resistance and tolerance in biofilms clinically and for molecules screening in anti-biofilm drug development. P. aeruginosa showed high levels of antibiotic tolerance, with minimal effect on bacterial load and increased proteases production, which could negatively affect lung function. This may potentially contribute to exacerbations and eventual lung failure.


Introduction
Cystic fibrosis (CF) is a genetic disease in which people have decreased mucociliary clearance in the respiratory tract, due to mutations in the cystic fibrosis transmembrane conductance regulator (CFTR) gene, which encodes a chloride channel [1,2]. This impairment leads to a reduction in mucus clearance and increased viscosity, resulting in accumulation of inhaled microbial cells, increased bacterial adherence and inflammation and the formation of bacterial biofilm [1][2][3]. Biofilm infections are more difficult to eradicate due the difference in their nature compared to non-biofilm infections; thus, they are lifelong infections in CF. These biofilm infections are characterised by acquiring distinctive resistance mechanisms compared with non-biofilm infections. There are three main mechanisms. First, there may be low antibiotic penetration into the biofilm due to the production of extracellular matrix. Second, the different bacterial metabolic states in the biofilm lead to increased phenotypic heterogeneity, affecting the success of treatment. Third, adaptive mechanisms controlling differential gene expression of multiple virulence factors, such as efflux pumps and antibiotic-degrading enzymes, lead to antibiotic tolerance [4,5]. The latter is highly dependent and varies based on the environment surrounding the biofilm [5]. In CF and other biofilm-based infections, antibiotic prescription is mainly based on standard minimum inhibitory concentration (MIC) methods, despite these being based on planktonic cells [6]. Planktonic-based diagnostics are suitable for detecting intrinsic and acquired stable resistance mechanisms; however, biofilm-based diagnostics will additionally detect environmentally-induced and biofilm-associated resistance mechanisms. There is a drastic increase in antibiotic tolerance using biofilm-based diagnostics, such as the Calgary device, in comparison to MIC methods, demonstrating the limitation of using planktonic-based models for biofilm infections [7]. Some of these in vitro biofilm models are robust for antibiotic susceptibility screening; they fail to recapitulate the complexity of biofilm infections, environment and host-dependent interactions [6,8], all of which affect the antibiotic susceptibility profile [8]. Also, these biofilm models have not been developed to demonstrate antibiotic susceptibility profile in multi-species infections. Thus, if diagnostic tests fail to accurately detect in vivo antibiotic resistance, this will result in in recurrent and complicated infections [8], and may lead to a vicious cycle of increased resistance.
In this study, we employed a previously developed ex-vivo pig lung biofilm model (EVPL) [9, 10] for antibiotic susceptibility testing of CF P. aeruginosa isolates and compared it with standard MIC and the Calgary device assays. The effect of exposure to antibiotics (ciprofloxacin, meropenem, tobramycin and combination therapy) on the virulence of P.
aeruginosa was also assessed. The results demonstrated an increased antibiotic tolerance in EVPL model at >25-fold the reported sputum concentrations when tested in Mueller-Hinton broth, which even further increased when tested in artificial sputum media. Exposure to antibiotics showed an increased production of total proteases, which may have a role in lung damage. Normalised proteases/cfu and pyocyanin/cfu demonstrated an increased production of these virulence factors per cell. Current clinical prognosis in CF is alarming and creates an urgent need to develop effective anti-biofilm agents. Thus, we propose a unique approach to P a g e 4 o f 1 9 predict the true clinical effect of antibiotic treatments on bacterial clearance and associated virulence factors in CF using the EVPL model.

1) Bacterial isolates
Clinical Pseudomonas aeruginosa isolates used in this study were isolated from a chronically-colonised CF adult, as described in Darch et al. [11]. Six bacterial isolates were chosen to represent a range of different growth rates and virulence profiles as reported by Darch et al. [11]. PA14 was used as a control laboratory strain for comparison. the absorbance was measured at 400 nm. PBS was used as a negative control and a standard curve using proteinase K ( Figure S1) was used to estimate the total amount of proteases.

2) Antibiotic susceptibility testing
Total pyocyanin was quantified according to Saha et al. [15] with minor modifications.
Briefly, pyocyanin was extracted using chloroform in a ratio of 5:3. The chloroform mixture was vortexed for 2 minutes, then centrifuged at room temperature at 10,000 rpm for 5 minutes. The bottom layer was transferred to a new 2 mL tube and an equal volume of 0.2M HCl was added. Tubes were vortexed for 2 minutes, centrifuged at room temperature at 10,000 rpm for 5 minutes and 200 µL of the top phase was transferred to a black 96-well plate and the absorbance was measured at 520 nm [15]. The concentration of pyocyanin (µg/mL) was calculated by multiplying the OD 520 by 17.072 [16].

5) Statistical analyses
All data were analysed by ANOVA to test for the main effect and interactions of different lung, strains and antibiotic treatments using RStudio v1.1.463 (©2009-2018 RStudio, Inc.).
Unpaired t-tests were performed for pairwise statistical analysis using GraphPad Prism (v8.0.1) and the familywise error rate correction for multiple comparisons.

1) Antibiotic susceptibility testing
Clinical P. aeruginosa strains were resistant to ciprofloxacin (MIC 2 µg/mL) and sensitive to meropenem (MIC ≤ 0.25 µg/mL) by standard antibiotic susceptibility testing using the broth dilution method; PA14 was sensitive to both antibiotics ( Table 1). Determination of MBECs using the Calgary device demonstrated an increase of 2-4-fold and 64-128-fold MIC for ciprofloxacin and meropenem, respectively (Table 1). Additionally, the MBEC recovery plates showed no visible production of pyocyanin, pyochelin or pyoverdine ( Figure S2).  Figure 1 represents a schematic diagram of the work flow as previously described. Figure S3 and Figure S4 show pieces of tissues infected with P. aeruginosa strains after 7 days of biofilm formation and the mucoid phenotype of the strains in the tissues, respectively, to represent chronic CF infection [17].  form biofilms and homogenised for the determination of the biofilm bacterial load. Replicate infected tissues were exposed to antibiotics for 24 hours before the decrease in bacterial load was determined.

2) Increased antibiotic tolerance of bacterial biofilms in the EVPL
The effect of antibiotics on the bacterial load was first tested by transferring bronchiole sections containing developed biofilms to MHB (standard medium for microdilution assays) containing antibiotics. This allowed to directly compare the inhibitory effect of these antibiotics in EVPL versus in standard diagnostics assays, without the effect of the medium or any physiological differences induced by CF lung mucus. Treatment with ciprofloxacin at 32-fold MIC (8-16-fold MBEC) resulted in a 1-2 log decrease in bacterial load across all tested clinical strains (Figure 2A). However, exposure to meropenem at 256-1024-fold MIC (4-8-fold MBEC) resulted in only about 1 log decrease of the bacterial load across all strains except SED 34, which showed less than a log decrease in the bacterial count ( Figure 2B).

3) The effect of antibiotic-mediated virulence
To showed increased pyocyanin secretion by 162% with exposure to meropenem ( Figure 4B and Table S2). Surprisingly, ciprofloxacin treatment also showed a significantly higher pyocyanin by 116% ( Figure 4B and Table S2). Proteases and pyocyanin concentrations in tissue and surrounding ASM, and total fold increases associated with antibiotic treatments are summarised in Table S1 and Table S2, respectively.
Because the total concentrations of proteases and pyocyanin discussed above are a function of both altered cellular production levels and altered cell numbers, we then normalised total proteases and pyocyanin concentrations by bacterial counts, to determine how antibiotic exposure affected per-cell production. The amount of proteases/cfu and pyocyanin/cfu measured in the tissues were slightly lower or equal to that of the surrounding ASM, respectively ( Figure 4C and 4D). Exposure to meropenem (64 µg/mL), ciprofloxacin (64 µg/mL) and tobramycin (4 µg/mL) slightly increased the production of proteases and pyocyanin by bacterial cfu, while a significant increase was found with tobramycin (200 µg/mL) and a combination of meropenem/tobramycin (64/200 µg/mL) (Figure 4C and 4D).
The latter treatments have a greater effect on bacterial load ( Figure S5). The total amount of tissue and ASM proteases/cfu and pyocyanin/cfu is shown in Figure S6.

Discussion
In CF, chronic P. aeruginosa infections are characterised by the mucoid phenotype, which can adversely affect the individuals' pulmonary function increasing mortality rates. . Therefore, we believe it is a more accurate model to show the effect of antibiotic treatment in CF is by testing for antibiotic susceptibility in ASM rather than general laboratory medium. This was alarmingly poorer than in MHB (Figure 3).
The work also indicated the potential in vivo effect of exposure to different antibiotics, in ASM, onto the virulence of P. aeruginosa to understand the clinical implications of chosen antibiotics. We focused on two main virulence factors of P. aeruginosa: proteases and pyocyanin production. Production of proteases is triggered by the quorum sensing system to degrade vital host proteins and antibodies. In CF lungs, proteases have been shown to cause a severe inflammatory response leading to pulmonary damage [24], and were detected in sputum during exacerbation [25]. Pyocyanin is also regulated by the quorum sensing system.
It is a redox molecule that generates reactive oxygen species to induce oxidative stress in host cells, leading to cell damage and lysis. P. aeruginosa is protected from these reactive oxygen species by its own catalases [24]. Previous studies had estimated the concentration of pyocyanin in sputum of as high as 16.5 µg/mL [26], similar to the detected values in Figure   4B. The antibiotic recovery plates of P. aeruginosa following Calgary biofilm susceptibility testing did not show any production of pyochelin, pyoverdine or pyocyanin (Figure S2), but following exposure to antibiotics in the EVPL model, bacterial expression of proteases and pyocyanin was increased (Figure 4). This highlights the aggressiveness of P. aeruginosa infection in CF.

Conflict of Interest
The authors declare no conflict of interest.  e  r  g  o  g  n  e  -B  e  r  e  z  i  n  E  ,  M  u  l  l  e  r  -S  e  r  i  e  y  s  C  ,  A  u  b  i  e  r  M  ,  D  o  m  b  r  e  t  M  C  .  C  o  n  c  e  n  t  r  a  t  i  o  n  o  f  m  e  r  o  p  e  n  e  m  i  n   s  e  r  u  m  a  n  d  i  n  b  r  o  n  c  h  i  a  l  s  e  c  r  e  t  i  o  n  s  i  n  p  a  t  i  e  n  t  s  u  n  d  e  r  g  o  i  n  g  f  i  b  r  e  o  p  t  i  c  b  r  o  n  c  h  o  s  c  o  p