A labelled-ubiquicidin antimicrobial peptide for immediate in situ optical detection of live bacteria in human alveolar lung tissue† †Electronic supplementary information (ESI) available: Experimental details and Fig. S1–S5. See DOI: 10.1039/c5sc00960j

A fluorescently labelled ubiquicidin peptide enables bacterial detection in human lung tissue in vitro.


Introduction
It is a global priority to reduce antibiotic resistance, 1 and consequently novel methodologies to rapidly diagnose or exclude bacterial infection are urgently required to implement antibiotic stewardship. Current diagnostic approaches rely on detecting the non-specic clinical features of a developing infection alongside optimal sampling of an affected area and subsequent bacterial culture. However, these approaches are slow, prone to contamination and diagnose infection at a late stage. Hence, rapid, sensitive and in situ molecular imaging assays for bacterial infections address an unmet need for bacterial diagnostics 2 and have the potential to lead to more effective antibiotic prescribing.
Pneumonia is an important clinical disease in which to develop and apply bacterial diagnostics. It is the leading infectious cause of mortality in children, 3 adults admitted to hospital with pneumonia have a mortality of up to 14%, 4 and in critically ill individuals in the intensive care unit (ICU), those who develop ventilator associated pneumonia (VAP) can have an attributable mortality of over 70% with some etiologies. 5 During the development of pneumonia, bacteria proliferate within the distal gas exchanging units of the human lung, an anatomical region in humans that is considered to be relatively sterile at basal microbiota levels. This is in contrast to other areas of the human body where bacterial burden from colonisation is high, such as the upper airway and gut, where distinguishing colonisation from pathogenic bacteria is a diagnostic challenge. Indeed current methodologies to diagnose pneumonia rely upon general sampling from upper airways and are prone to contamination and consequent 'false-positive' initiation of antibiotic therapy.
Optical molecular imaging (OMI), is a relatively novel diagnostic approach that has utility in preclinical and clinical applications. 6 Advantages of OMI include: (i) real-time imaging, (ii) high resolution, (iii) absence of radiation-related risks and (iv) low costs. However, whilst the applications of molecular imaging are showing increasing utility in disease areas such as oncology, there are a limited number of functional effective molecular imaging approaches for bacterial detection. 7 A rational and accepted approach for targeted bacterial molecular imaging has been to label antimicrobial peptides (AMPs), widespread components of the innate immune systems of virtually all multicellular organisms. 8 AMPs possess a number of factors making them promising tools for bacterial OMI including their amino acid composition, amphipathicity and cationic charge that enable selective insertion and binding to the bacterial membrane. 9 The most studied AMP for infection imaging is the 59 amino acid cationic AMP, ubiquicidin (UBI) and in particular a UBI fragment containing thirteen amino acid residues (UBI 29-41 ) [Thr-Gly-Arg-Ala-Lys-Arg-Arg-Met-Gln-Tyr-Asn-Arg-Arg] which has been utilised in radionucleotide based imaging. 10 However its use involves unavoidable radiation exposure and is not suitable where portable solutions are required, such as critically ill ventilated patients with suspected pneumonia. More recently a near-infrared labelled version of UBI 29-41 was applied in a murine model of bacterial infection. 11 Unfortunately, the synthetic route led to a multitude of products with the dye dominating the distribution of the construct, resulting in a similar distribution for the dye alone compared to the uorescently labelled UBI 29-41 . In addition there are obvious "Achilles heels" in a labelled UBI 29-41 peptide with respect to potential degradation in vivo, exacerbated even more so in sites of active inammation by proteolysis and oxidation. Building on these developments and also in recognition of the feasibility of developing an OMI agent (Smartprobe) based upon a UBI scaffold, we embarked upon a lead optimisation approach to yield a Smartprobe suitable for application in diagnosing pneumonia (bacterial infection in the distal lung) when partnered with an optical imaging device. For this Smartprobe to be valuable bio-medically we optimised a series of prerequisites including; (i) Specicity and selectivity to label clinically relevant bacteria, (ii) High signal-to-noise contrast,  (iii) Resistance to degradation and stability in the presence of human lung bronchoalveolar lavage uid (BALF) from patients with acute respiratory distress syndrome (ARDS), (iv) High affinity and bacterial detection by a clinically approved optical imaging device in human lung alveolar tissue.
We demonstrate the development and optimisation of uorescently labelled UBI 29-41 , through sequential modication of the uorophore label, peptide sequence and secondary structure, and demonstrate the ability of the lead compound to detect live bacteria in human lung tissue.

Results and discussion
We synthesised a series of UBI peptides to assess structure and functional consequences for bacterial labelling (Table 1) and to remove proteolytic susceptibilities.  7-Nitrobenz-2-oxa-1,3-diazole (NBD) demonstrates increased signal to noise over carboxyuorescein (FAM) UBI conjugates UBI 29-41 was rstly conjugated to carboxyuorescein (UBI-1) and used to image bacteria with analysis via confocal microscopy. This yielded insufficient signal to noise to distinguish bacteria from background uorescence (Fig. 1). We therefore took advantage of the environmental reporting properties of the uorophore 7-nitrobenz-2-oxa-1,3-diazole (NBD) 12 and hypothesized there would be uorescence amplication of NBD in the apolar hydrophobic bacterial membrane. UBI-2 demonstrated good signal to noise and was able to detect bacteria (Fig. 1).
Removing and replacing the methionine residue improves stability UBI-2 underwent chemical degradation upon storage in phosphate buffered saline at 37 C, which was prevented by substitution of the Met to Nle, giving rise to UBI-3 (Fig. 2). This modication is particularly of relevance due to the expected in vivo inammatory oxidative environment within an infected pneumonic lung due to inltrating host inammatory cells such as neutrophils, 13 and therefore all subsequent modications incorporated Nle in place of Met.

Controlling degradation in human bronchoalveolar lavage uid (BALF)
To determine the ability of UBI-3 to image bacteria in the continued presence of a complex inammatory environment, bacteria were labelled with UBI-3 in the presence of bronchoalveolar lavage uid (BALF) retrieved from ventilated ICU patients with Acute Respiratory Distress Syndrome (ARDS). ARDS BALF was utilised as ARDS is a severe life threatening pulmonary syndrome, characterized by a massive proteolytic and oxidative inammatory environment within the distal alveolar region. 14 Under these harsh experimental conditions, there was a signicant reduction in the uorescent labelling intensity of bacteria with UBI-3 (Fig. 4). This was due to the rapid (<5 min) degradation of the peptide as demonstrated by Matrix-Assisted Laser Desorption/Ionization Time-of-Flight Mass Spectrometry (MALDI-TOF MS) analysis (ESI Fig. S1 †). To design out susceptible amino acid cleavage points, a number of modications were made (Table  1) and the functional consequences assessed. The chemical approaches pursued to improve resistance to degradation included replacing amino acids at the cleavage positions with non-natural amino acids such as N-methyl or D-amino acids, 15 and cyclisation of the peptide, 16 which could have the added benet of enhancing affinity towards bacterial membrane. 17

Modication of mass spectrometry identied degradation sites
The Arg-Arg bonds were identied as the prominent position of degradation (Fig. 5) and Compounds UBI-4-9 (Table 1) were synthesised with a variety of substitutions around the degradation sites. The susceptible peptide bonds at P1-P2 and P7-P8 were blocked using N-MeArg (UBI-4), and to further mask these positions, a long polyethylene glycol (PEG) chain and/or a D-amino acid were positioned at the carboxylic and/or amino termini (UBI-5-7). N-MeArg insertion (UBI-4) or N-MeArg with two D-amino acid residues at the C-terminal (UBI-5) improved stability in ARDS BALF (ESI Fig. S2 †) and allowed bacterial labelling at the same intensity as UBI-3 (Fig. 6). However, insertion of a PEG chain at both termini (UBI-6) and constructs with entirely D-amino acids (UBI-8 and UBI-9), whilst improving stability, resulted in loss of labelling on bacteria (ESI Fig. S2 and 6 †). A single PEG at the C-terminus alone (UBI-7) did not improve stability in ARDS BALF, and resulted in loss of bacterial labelling.

Cyclisation of the peptide sequence
We hypothesized that cyclisation of the sequence would allow for greater stability and also improve bacterial affinity and therefore cyclic variants of UBI-3 were synthesized (UBI-10 and UBI-11), which were both found to be stable in BALF for 5 minutes (Fig. 7 and S3 ESI †). Compared to the UBI-3 there was a signicantly higher signal for equivalent concentrations of the cyclic construct (UBI-10) (Fig. 8). However, these studies also highlighted the degradation of UBI-10 at 10 minutes, guiding the temporal constraints of Smartprobe delivery and optical imaging. UBI-11 demonstrated that a single amino acid change imparted stability in BALF for over 30 minutes and the compound showed comparable bacterial labelling to UBI-3, (ESI Fig. S3 †) however the enhanced bacterial labelling seen with UBI-10 was lost (Fig. 8).

Bacterial affinity evaluation
Three compounds (UBI-3, UBI-5 and UBI-10) were assessed for further evaluation. The rationale for assessing these variants was that UBI-3 and UBI-10 were native UBI sequences in linear or cyclic secondary structures respectively and UBI-5 was stable and demonstrated a similar bacterial labelling to UBI-3. Bacterial labelling was demonstrated in a concentration dependent manner for all three compounds. The relative uorescence of UBI-10 labelled bacteria was higher than UBI-3 or UBI-5 labelled bacteria (Fig. 9). At 10 mM, UBI-10 retained bacterial specicity over neutrophils. However at 50 mM, some neutrophil labelling was observed (Fig. 9). In vitro affinity of bacterial binding was assessed by incorporating a wash step in the live confocal imaging protocol with a panel of bacteria with the 'wash-off' representing a surrogate indicator of Smartprobe-bacterial labelling affinity. This demonstrated that following labelling and a wash, there was signicantly lower uorescence for the linear compounds (UBI-3 and UBI-5) than the cyclic compound without modication (UBI-10) (ESI Fig. S4 †) against three different bacterial species. UBI-10, therefore, was assessed further in a clinically relevant biological system.
Bacterial detection by a clinical bre-based optical imaging in human lung tissue Fibered confocal uorescence microscopy (FCFM) with a 488 nm laser system enables in vivo cellular and subcellular resolution imaging deep in human lung. 18 This system allows imaging of the intrinsic autouorescence of elastin thereby enabling alveolar imaging. Therefore, to assess whether this platform has the resolution to detect uorescent bacteria, we pre-labeled bacteria (with calcein AM) and then demonstrated that we were able to detect uorescently labeled bacteria on the background of human lung autouorescence (ESI Fig. S5 †).
Since NBD is also excited at 488 nm, the utility and detection of the Smartprobes with FCFM was next evaluated. Bacteria were labelled with UBI-3 and UBI-10 and imaged in suspension demonstrating the punctate uorescence pattern of labelled bacteria. These were then delivered onto ex vivo human lung tissue fragments. Bacteria labelled with UBI-3 immediately lost their uorescent signal whereas UBI-10 labelled bacteria, were readily identied above elastin auto-uorescence (Fig. 10).

Summary
We synthesised a library of uorescently labelled UBI 29-41 mimetic peptides for rapid, live, bacterial imaging; with the ambition to develop Smartprobes that demonstrate potential clinical utility in conditions of high unmet need; such as the rapid diagnosis of suspected pneumonia in critically ill individuals. We discovered key determinants of peptide stability and bacterial binding affinity as potential obstacles for clinical utility. A number of structural modications were subsequently engineered to drive improved stability, resistance to degradation and improve function. A cyclic variant of NBD-UBI (UBI-10) proved to be the most functional and specic Smartprobe for bacterial detection and showed proof of concept for detection of bacteria over intrinsic lung elastin autouorescence in human lung tissue.
These ndings provide pivotal insights into future strategies to develop chemical Smartprobes capable to detect bacteria within the distal human lung.