Using fluorescence microscopy to shed light on the mechanisms of antimicrobial peptides

Abstract

Antimicrobial peptides (AMPs) are promising in the fight against increasing bacterial resistance, but the development of AMPs with enhanced activity requires a thorough understanding of their mechanisms of action. Fluorescence microscopy is one of the most flexible and effective tools to characterize AMPs, particularly in its ability to measure the membrane interactions and cellular localization of peptides. Recent advances have increased the scope of research questions that can be addressed via microscopy through improving spatial and temporal resolution. Unique combinations of fluorescent labels and dyes can simultaneously consider different aspects of peptide–membrane interaction mechanisms. This review emphasizes the central role that fluorescence microscopy will continue to play in the interrogation of AMP structure–function relationships and the engineering of more potent peptides.

Two concerning trends have arisen in therapeutic approaches to bacterial infections. First, reports have noted the increasing prevalence of bacterial strains that are highly resistant to conventional antibiotics [1]. Most concerning are strains that harbor resistance to multiple antibiotics and organisms that show resistance to antibiotics once considered to be drugs ‘last resort,’ such as vancomycin. Antibiotic-resistant infections are estimated to annually cause over 23,000 deaths in the USA [1] and 33,000 deaths in the European Union [2], in addition to increasing recovery time and disability for substantially larger populations. There is also a significant financial impact of resistant infections, as the annual medical cost of treating antimicrobial resistance is upwards of $2 billion in the USA alone [3,4]. Alongside this increased resistance, the number of newly developed antibiotics approved for clinical use dramatically decreased from the 1980s through the early 2010s [1]. This decrease was coupled with a reduction in major pharmaceutical companies actively pursuing antibiotic research and development. This decline has led to concerns that the pipeline of new antibiotics will not be sufficient to address increasing concerns about resistance.

Antimicrobial peptides (AMPs) have emerged as a potentially valuable addition to the therapeutic arsenal against bacterial infections [5]. AMPs are naturally produced as part of the innate immune system in a wide variety of organisms, including plants, animals (including humans) and fungi. Even bacteria produce AMPs as a means of combating other bacterial strains. The majority of AMPs isolated from natural systems are relatively short (e.g., 10–30 amino acids long), cationic peptides. The positive charge on these peptides is believed to promote their interactions with bacteria, which typically have more negatively charged membranes and a more negative transmembrane potential than eukaryotic cells [6]. Importantly, although bacterial resistance to AMPs does occur [7,8], it does not appear to arise as readily as resistance to other antibiotics. Researchers hypothesize that this reduced resistance is due to AMPs frequently having multiple targets, the challenges in bacteria modifying membrane structures, and the overall polyanionic nature of many prokaryotic macromolecules [5,8–10]. This potential ability to better elude resistance elevates the potential of AMPs to help address the increasing need for new antibiotics.

For decades, researchers have effectively isolated AMPs from natural sources. Isolation studies typically involve identifying a potential source and performing methodical and thorough purification to identify the active natural peptide. As these studies have identified over 3000 naturally occurring AMPs [11], it has become increasingly important to investigate how these AMPs halt the growth of and/or kill bacteria. This mechanistic understanding is critical to efforts to optimize and refine the activity of AMPs to make them most effective for clinical purposes.

As membranes provide the primary barrier between bacteria (target cells) and the external environment, investigating the means by which AMPs interact with bacterial membranes highlights an early step within their overall mechanism(s) of action. This review will focus on the use of fluorescence microscopy in examining the mechanisms of peptide-membrane interactions. The ability to image the localization and effect of AMPs as they encounter and attack bacteria is an essential part of determining their overall mechanisms of action. In addition to utilizing instrumentation generally accessible to researchers at a wide variety of institutions in relatively broad resource settings, fluorescence microscopy approaches allow for measurements of cellular activity and membrane properties under physiological and dynamic conditions not possible with electron microscopy. While not the focus of this review, we do want to note the contributions made by recent applications of atomic force microscopy (AFM) on AMP systems, particularly those probing the effect of AMPs on membrane structures [12,13].

Fluorescence microscopy applications to analyze AMP interactions with bacterial membranes & entry into cells

Antimicrobial peptides are traditionally believed to follow one of two general categories of lipid interactions in early steps of their antimicrobial mechanism [5,6,8,14]. One set of mechanisms involve peptides that localize to the cell membrane. In many cases, these membrane localized peptides induce permeabilization, in which the peptide disrupts the bacterial membrane in such a way that ultimately kills the cell. This can occur through pore formation or more generic disruption, such as that described by the carpet mechanism. However, in other cases, a membrane localized peptide can harm the bacterium through a nonpermeabilizing mechanism, such as interactions with cell–wall precursors. The other common process is translocation, in which peptides move across the membrane without causing significant disruption and then interact with internal targets, such as DNA. Although there is evidence that peptides can engage in more than one type of interaction [8,14,15], generally antimicrobial peptides are categorized as acting primarily by one type of membrane mechanism.

Fluorescence microscopy can be a particularly valuable tool for helping to determine whether an AMP appears to follow a primarily membrane localizing or translocating pathway by visualizing whether the peptide is confined to the bacterial membrane or enters the cell. A clear advantage of microscopy is the ability to consider the effect of a peptide on the single-cell level. However, in order to effectively determine whether a peptide has entered a single cell or remains membrane localized, it is advantageous to image ‘slices’ through the whole bacterial cell, often called a ‘z-stack’ (Figure 1). This is necessary because, given the micrometer scale of a typical bacterium, a single, widefield, epifluorescence microscopy image could be taken such that it largely includes not only the cytoplasmic space but also the membrane on ‘top’ or ‘bottom’ of the cell. In this case, a peptide that actually is localized to the membrane could look like it has translocated into and spread throughout the cell interior. Because confocal microscopy allows the user to focus excitation light and allows detection of fluorescence only from a narrow width of the cellular sample, it leads to the observation of less ‘stray’ fluorescence from outside the focal point/slice than would be visible using a widefield approach. For studies focused on membrane localization, total internal reflection fluorescence (TIRF) microscopy offers additional clear advantages in resolution and signal to noise, as it only excites molecules within 100–200 nm of the surface of the cell. To image deeper into the cytoplasmic space of a bacterium, highly inclined and laminated optical sheet microscopy might be necessary and, again, has the advantage of a high signal-to-noise ratio [16–18]. However, confocal microscopy's ability to maneuver the position of imaging from one surface through the cell interior and then across the opposite surface may provide the most versatile approach, allowing better resolution of both ‘membrane’ and ‘internal’ slices.

Figure 1. . Confocal microscopy allows for researchers to image axial ‘slices’ through a bacterial cell, often called a ‘z-stack’.

Collecting a z-stack makes it possible to distinguish whether a fluorescent signal is only located on the membrane or throughout the cell. This image shows a set of z-stack images from an E. coli cell incubated with a fluorescently labeled peptide that is localized to the membrane. Collecting only a single image representing the ‘top’ or ‘bottom’ of the cell could lead a researcher to incorrectly conclude that a peptide was observed throughout the cell.

An excellent example of the utility of fluorescence microscopy imaging in measuring cellular localization in order to provide information on AMP mechanisms of action is the characterization of histone-derived antimicrobial peptides (HDAPs) by several researchers. Over the past few decades, researchers have isolated a number of AMPs from natural sources that are fragments of histone proteins [19,20]. The best studied of these HDAPs is BF2, a peptide derived from the naturally occurring buforin I peptide from the toad Bufo bufo gargarizans [21]. Early characterizations of BF2, including confocal fluorescence microscopy imaging, highlighted its ability to translocate into bacterial cells [22,23]. Moreover, imaging of a series of BF2 mutants identified its single proline residue as critical in its ability to cross membranes [22].

Based on these and other studies, our lab designed a series of novel HDAPs sharing structural features important for BF2 function [24]. Confocal fluorescence microscopy measurements were critical in identifying that the most active peptides had dramatically different patterns of cellular localization, as DesHDAP1 translocated into bacteria akin to BF2 while DesHDAP3 localized to and disrupted the cell membrane, suggesting a permeabilizing membrane interaction [25]. Confocal fluorescence microscopy has also helped to dissect the function of hipposin, an intriguing naturally occurring peptide that contained the sequence of two HDAPs, parasin and BF2, as well as a previously unidentified sequence named HipC [26]. This work showed how combining BF2 (a translocating peptide) with parasin [27], which localizes to membranes, produces a hybrid peptide that is membrane localizing. These data also showed that HipC is able to effectively enter cells despite not having any measurable antimicrobial properties [26]. While the studies mentioned above focused on qualitative analysis of bacterial images (i.e., there is fluorescence observed inside a cell or only around the membrane), one can leverage the ability to quantitate the intensity of fluorescence observed via confocal microscopy. For example, it is possible to define cellular entry of peptides by quantitatively comparing the fluorescence intensity in regions inside a cell to the intensities in the membrane region and outside the cell [28].

However, even with confocal microscopy, the small size of bacterial cells, with dimensions typically on the order of 1 μm, can make it difficult to resolve the difference in fluorescence at the membrane versus the inside of the cell, as mentioned above. For example, the practical optical, axial (z-direction) resolution limit for confocal configurations (with high numerical aperture objectives and short excitation wavelengths) is about 0.4 μm. Thus, even an ‘internal’ bacterial slice could contain some membrane fluorescence (Figure 2A). Because of this light resolution limit, researchers who want to capitalize on the advantages of fluorescence microscopy have turned to approaches that allow them to either modify or model the cellular system to increase spatial resolution. These systems can also be employed in conjunction with cell chambers that allow for single-cell measurements with enhanced time resolution.

Figure 2. . Optical resolution considerations for z-stack ‘slices’.

(A) The optical resolution of slices taken in confocal microscopy is typically limited to 0.4–0.8 μm, depending on the wavelength of emission, instrument optics and other settings. Thus, it can be difficult to obtain ‘slice’ that does not collect at least some membrane fluorescence when imaging standard bacteria. (B) The increased size and morphological symmetry of bacterial spheroplasts and protoplasts make it more feasible to collect slices that contain only intracellular fluorescence. A similar effect is also observed in the imaging of spherical giant unilamellar vesicles.

Modified cells & lipid vesicles can help address resolution challenges in imaging bacterial cells

One approach to addressing resolution issues has been for researchers to use ‘modified cells’ that have more favorable size and/or morphology for imaging. To this end, our labs have pioneered the use of bacterial spheroplasts and protoplasts for confocal imaging of AMPs. Spheroplasts and protoplasts are bacteria cells that have been grown under conditions that allow them to form larger, spherical shapes with diameters typically in the range of 2–5 μm (Figure 3) [28–32]. For spheroplasts, this is done by exposing a Gram-negative bacterial culture to an antibiotic, such as cephalexin, that allows cells to continue growing but not divide. This leads to the formation of large ‘snakes’, that can then be exposed to lysozyme to disrupt the outer membrane and form the spheroplast [29,30,33]. Although modified, spheroplasts remain viable as they can be returned to a normal culture if grown under appropriate conditions [33,34]. Protoplasts are made by analogous methods from Gram-positive strains [31,35]. The size and morphological symmetry of spheroplasts and protoplasts makes it easier to distinguish between membrane and cytoplasm fluorescence than with unaltered bacteria where the random orientation of each individual cell in the sample limits the capture of z-slice images in which membrane and cytoplasm are definitively distinct (Figure 2). Using spheroplasts has allowed us to further study whether different HDAPs are able to effectively translocate across bacterial membranes (Figure 4) [15,36]. This approach also allows the collection of substantially larger numbers of images than when using intact bacteria since the larger and spherical altered cells are more facile to locate and image. The ability to acquire robust datasets and the enhanced spatial resolution allow for more thorough and systematic characterization of AMP interactions with membranes.

Figure 3. . Confocal microscopy images of Escherichia coli at different stages during the formation of spheroplasts, highlighting the size and morphology differences of the ‘snake’ and spheroplast forms produced.

Fluorescence shown is from the membrane dye di-8-ANEPPS.

Adapted with permission from [15].

Figure 4. . Confocal images of Escherichia coli spheroplasts incubated with FITC labeled versions of the buforin II and buforin II P11A peptides.

The merged fluorescence of the peptide FITC label (green) and di-8-ANEPPS (red) used to label the bacterial membrane is shown. The relative localization patterns of the peptides are emphasized through the collection of z-stacks. Images from three different positions in a single z-stack are shown for each peptide; z-positions are given relative to the middle image of each stack.

FITC: Fluorescein isothiocyanate.

Adapted with permission from [15].

Other researchers have used spheroplasts to investigate membrane leakage caused by AMPs. Huang and coworkers examined the leakage of dye into spheroplasts caused by the presence of melittin and LL37 [37]. The altered morphology of spheroplasts enhanced the feasibility of imaging changes in a single cell over time in response to added peptide, including through the use of a fluorescence recovery after photobleaching analysis. This approach demonstrated that the flux of dye molecules in and out of the spheroplasts was consistent with diffusion through pores formed by membrane localized AMPs.

Another approach to creating larger bodies for imaging utilizes lipid vesicles. These vesicles can be made with a wide variety of different lipid compositions, and various preparation protocols can be used to tailor their size [38,39]. Bulk assays that measure the fluorescence of a population of vesicles using a fluorescence spectrometer have been widely applied to assess the ability of AMPs to disrupt or translocate across membranes [38,40,41]. Although bulk assays typically utilize small or large unilamellar vesicles, giant unilamellar vesicles are typically preferred for microscopic imagining due their larger size, on the order of micrometers.

The utility of imaging GUVs is demonstrated by a series of studies characterizing the peptide daptomycin. The Huang lab used confocal microscopy to visualize the colocalization of lipid and daptomycin aggregates and to quantify how peptides alter the area and volume of GUVs [42]. These membrane area measurements required significant precision as they typically represented changes of <4%, a change that would be challenging to capture in a living cell. In another study, the Almeida lab looked at the interaction of daptomycin using GUVs consisting of phosphatidylcholine and phosphatidylglycerol [43]. They used confocal microscopy to monitor the binding of daptomycin to GUVs and the resulting effects on the vesicle membrane. Notably, their approach permitted the collection of images from over 1000 GUVs, allowing for robust data analysis. These quantitative, colocalization analyses showed that daptomycin does not form a pore but does cause the clustering of phosphatidylglycerol lipids in the membrane, as seen through the change in average fluorescence around the GUVs arc. They suggested that the clustering of lipids is the formation of daptomycin-POPG domains in the membrane that eventually causes GUV collapse, providing further insight into daptomycin's mechanism of antimicrobial action.

The Almeida lab has also used GUVs to consider the ability of AMPs to translocate across cell membranes. Their approach to measuring translocation was based on the observation that their vesicle preparation created some GUVs that contained inner vesicles [44,45]. Thus, by using confocal microscopy, they could monitor the ability of an AMP to bind to the surface and/or disrupt the integrity of this inner vesicle, which could only occur if the peptide was first able to translocate across the main GUV membrane. Their imaging approach with GUVs has also allowed them to visualize the flux of fluorescent molecules into vesicles to assess the effect of the peptide on membrane permeabilization [46], allowing for thorough analyses of peptide function [47].

Another example of GUV imaging is described by Moniruzzaman et al. in their characterization of LfcinB (4–9), a shortened derivative of lactoferrin B [48]. In this study, researchers utilized a combination of fluorescently labeled peptide and an encapsulated fluorophore to monitor the ability of LfcinB (4–9) to cause membrane leakage or enter either E. coli cells or GUVs using confocal microscopy. This study showed that LfcinB (4–9) can translocate across a membrane in both systems without causing leakage, suggesting that its antimicrobial activity is not due to membrane damage. The authors were able to show a series of analyses with GUV data considering the kinetics of peptide entry into vesicles. Moreover, the use of GUVs also emphasized that LfcinB translocation was not dependent on proteins or other nonlipid components of the bacterial membrane.

Notably, the microscopy methods developed to study the membrane interactions and internalization of AMPs can also be applied to the characterization of cell-penetrating peptides (CPPs), which share many similar features but do not themselves affect the physiology of the target cell [49]. For example, Almeida and coworkers have applied their GUV system to consider the membrane interactions and translocation of both AMPs and CPPs [44–46]. Similarly, Moghal et al. were able to use their approach to analyze the ability of the CPP transportan 10 to cross lipid membranes without inducing membrane damage on the single vesicle level [50].

Some work has compared the effect of AMPs on E. coli spheroplasts and GUVs [51]. In this study, no difference was found for steady-state membrane permeabilization by AMPs in GUVs compared with spheroplasts. The primary difference found was that spheroplast membranes responded more similarly to a wider variety of AMP concentrations, while GUVs showed different behavior at concentrations outside of a narrower range that matched that of spheroplasts. The authors partially ascribe these and some other details in the response to peptides to differences in the inherent tension of spheroplast and GUV membranes. Nonetheless, the researchers concluded that both systems were comparable model membranes to use when investigating the effects of AMP.

Despite their utility for overcoming issues of resolution in imaging bacterial cells, both spheroplasts and vesicles do require the use of systems that are not entirely physiological. In spheroplasts, the lack of an outer membrane means that peptides that function via interactions with cell–wall components cannot be effectively considered, although comparisons of spheroplasts to ‘normal’ cells could help to elucidate these cell–wall interactions. Additionally, when progressing to spheroplast form, cells may adjust other, potentially uncharacterized, aspects of their physiology. The significant changes in morphology between a rod-shaped cell and spherical spheroplast also could impact peptide mechanisms dependent on or regulated by membrane curvature. That said, the observations that spheroplasts remain viable and maintain some physiological functions of bacteria provide support for their use [33,34]. Moreover, spheroplasts allow researchers to use a membrane composition closely analogous to that of the native bacteria, including nonlipid components, which is not possible with vesicle model systems. That said, the ability to dictate the lipid composition of vesicles can allow researchers to consider questions related to the role of specific lipids, and AMP behavior with vesicles typically mirrors that observed with bacterial cells [38].

Using multiple fluorescent labels can allow for imaging sublocalization of peptides or simultaneously measuring different mechanistic aspects of peptides

As discussed above, visualization of peptide localization within a bacteria cell with fluorescence microscopy is frequently used to help determine a peptide's overall membrane interaction mechanism. Images of fluorescently labeled peptide, even if taken as confocal z-slices, may be best suited to address the binary question of whether a peptide has or has not entered a cell. Additionally, when only the peptide is labeled, researcher bias becomes a concern in selecting imaged cells to analyze because cells where peptide has translocated into the cytoplasm are brighter and more likely to be detected (by both the imaging hardware/software and operator; Figure 5). Thus, our lab has recently adopted the use of membrane dyes, such as di-8-ANEPPS, along with FITC (fluorescein)-labeled peptides in order to more clearly delineate whether peptides have localized to the membrane or effectively translocated into the cell [15,28,36]. Incorporating a fluorescent membrane label also allows one to image all the cells in a population, not only the ones that are interacting with fluorescently-labeled peptide. This approach has allowed for more quantitative co-localization analyses [28,36], and the presence of membrane dye can also help to monitor membrane morphology.

Figure 5. . Confocal images of Escherichia coli spheroplasts incubated with fluorescein isothiocyanate labeled peptides.

Images are shown from the middle image of a z-stack. (A) If spheroplasts are identified by viewing the channel with peptide fluorescence (green), it is possible for a researcher to be biased toward imaging cells demonstrating translocation due to the enhanced green color throughout the cell. (B) If spheroplasts are identified by viewing the channel with a membrane dye fluorescence (here di-8-ANEPPS, shown as red), there is reduced bias toward imaging cells showing membrane translocation.

However, a peptide could potentially localize to the membrane in fluorescent images but not actually disrupt the membrane. Thus, analysis of imaging data alone could lead to incomplete (or even erroneous) conclusions about the membrane interaction if one assumes that membrane localization always corresponds to membrane permeabilization (Figure 6). Typically, imaging data is accompanied by separate bulk experiments to verify membrane disruption and access to intracellular spaces, such as in our studies of HDAPs [26,40]. This is often accomplished using absorbance or fluorescence spectrophotometry to monitor cellular assays employing a colorimetric or fluorescent reporter, respectively. Propidium iodide, a fluorescent nucleic acid intercalator, which is impermeant to intact cells, is a classic example [52]. Modern microscopes permit experimental designs that incorporate simultaneous imaging of multiple fluorophores, allowing for the concurrent measurement of a labeled peptide and another fluorescent probe, like propidium iodide (Figure 6). These studies can provide information about peptide subcellular localization and membrane interaction mechanism alongside other aspects of antimicrobial mechanisms.

Figure 6. . Dual labeling offers advantages toward determining the mechanisms of antimicrobial peptide interaction with bacterial membranes.

This example shows anticipated results with a FITC-labeled AMP (green) in combination with the membrane impermeant nucleic acid intercalator propidium iodide (red). Top left: AMP localizes to the membrane and also permeabilizes the membrane allowing propidium iodide to enter the cell, bind to nucleic acids and fluoresce. Top right: AMP both permeabilizes the membrane and translocates into the cytoplasm. Yellow represents the combination of fluorescence from AMP and propidium iodide. Bottom left: AMP localizes to the membrane but does not disrupt membrane integrity leading to cells outlined in green. Bottom right: AMP translocates but leaves the membrane integrity intact leading to diffuse green fluorescence in the cytoplasm.

AMP: Antimicrobial peptide; PI: Propidium iodide.

As a further example, functional data about membrane integrity can be obtained when fluorescently-labeled AMPs are used in combination with small molecule probes. Vesicles can be prepared with membrane-impermeant fluorescent labels trapped in their interior such that the loss of intensity inside the vesicle indicates membrane breach. Alternatively, cellular studies can employ probes that are themselves reactive and capable of transferring a fluorophore to intracellular target or probes that rely on intracellular reactions or enzyme catalysis. The former, such as AlexaFluor hydrazides, are cell-impermeant, so only when the membrane integrity is compromised do they render intracellular molecules fluorescent [52]. The latter are typically cell-permeant probes that only fluoresce once the parent molecule is altered by an intracellular reaction, such as the cleavage of calcein AM by constitutive, intracellular esterases [52]. Complex and creative reporter systems using small molecule probes from the Yamazaki lab have demonstrated the translocation of LfcinB (4–9) in both GUVs and E. coli without compromising the membrane [48].

Other modern dyes allow researchers to monitor structural properties of membranes in imaging experiments. For example, the fluorescence emission of laurdan can be used to assess the fluidity of lipid membranes. A recent study probing the mechanism of cWFW considered images that used peptide with either Nile red, a general membrane stain, or laurdan [53]. Images of Nile red and fluorescently labeled cWFW emphasized how membrane domains containing peptide formed in Bacillus subtilis membranes after peptide exposure. Other images of cells exposed to cWFW and labeled with laurdan demonstrated the decreased fluidity of those regions. Similarly, other studies have effectively used laurdan to note the effect of other AMPs such as daptomycin, tyrocidines, and gramicidin S on membrane fluidity [54,55].

Beyond simply assessing membrane integrity, structure, and/or dynamics the potential targets for translocating peptides, such as nucleic acids, can also be investigated using nucleic acid-binding fluorophores (e.g. propidium iodide, SYTOX green, DAPI) in conjunction with fluorescently-labeled AMPs. For example, Hayden et al. examined bacterial nucleoids as a possible intracellular target of the AMPs piscidin1 and piscidin 3 [56]. The Gram-negative E. coli and Gram-positive Bacillus megaterium were combined with the nucleic acid dye DAPI and the FITC-labeled piscidins before examination via confocal microscopy and colocalization analyses. Though piscidin 1 was found to have more pronounced membrane leakage than piscidin 3 via membrane-leakage assays using both fluorescent and colorimetric methods, both translocated toward the nucleoids in E. coli and B. megaterium.

Using multiple fluorophores can also offer advantages even when the AMP itself is not labeled. For example, the bacterial cytological profiling approach introduced by Pogliano and coworkers utilized images taken with the membrane dye FM4-64 and DAPI as a means of characterizing morphological changes upon antibiotic exposure [57]. These changes could then be associated with one of five major classes of antibiotic mechanisms of action. In addition to its use for small molecule antibiotics, this approach has also been applied to AMPs [58]. In another example, striking images showing how acyldepsipeptides prevent cell division in Gram-positive bacteria were produced in cells with a membrane label and where candidate, key proteins were tagged with green fluorescent protein (GFP). Via fluorescence microscopy, Sass et al. elegantly showed that acyldepsipeptides inhibit septum formation due to the delocalization of several proteins critical to Z-ring formation including FtsZ and penicillin-binding protein 2 [59].

Fluorescent proteins, such as the classical GFP, have also been utilized to study the interaction of AMPs with cell membranes in addition to the use of small molecular fluorophores. The Weisshaar lab has adapted an E. coli system where GFP is solely expressed in the periplasmic space [60] to separately investigate peptide interaction with and permeabilization of the outer membrane and cytoplasmic membrane [9,61–64]. In a further study, the interaction of melittin with the membranes of single E. coli cells was assessed with both the periplasmic GFP reporter and the cell-impermeant probe SYTOX orange, which become fluorescent upon binding with intracellular DNA [65]. They observed in real time (through the use of cell chamber technology discussed below) that melittin first accumulated around the outer membrane. Then, with some delay, GFP escaped from the periplasmic space and the cells shrunk, indicating breach of the outer membrane. Rapidly, the cytoplasmic membrane deformed/blebbed and became permeable to SYTOX orange and then GFP. Unexpectedly, after both permeabilization events, the membranes resealed themselves for a period of time before developing a slow leakage from both membranes that eventually resulted in cell death. Visualizing the effects of melittin on E. coli using this creative approach allowed for the discovery of its previously unobserved sequence of antibacterial events. Moreover, in addition to the increased time resolution providing data from the initial stages of peptide/bacterial interactions (as discussed below), the clever use of SYTOX dyes in combination with periplasmic GFP provides subcellular localization information allowing researchers to infer mechanisms of action by drawing conclusion about where peptides localize and interfere with cellular processes.

As an example of an alternative approach with fluorescent proteins, Omardien et al. used GFP-tagged, membrane-bound proteins involved in essential cell processes in the Gram-positive B. subtilis to assess the downstream effects of the AMPs TC19, TC84 (derivatives of TC-1) and BP2 (a synthetic peptide based on human bactericidal permeability increasing protein, BPI) alongside SYTOX green membrane leakage experiments [66]. SYTOX green assays suggested the membrane is the primary target of the AMPs and their interactions then induce delocalization or abnormal localization of key proteins for cell envelope production and maintenance (among others) limiting their functionality and leading to adverse effects on cell viability.

One potential challenge to many of these visualization studies is that they employ fluorescently labeled peptides. Even relatively ‘small’ fluorophores, such as fluorescein- or rhodamine-derived molecules, are typically the size of one or two amino acids and could have a significant impact on the physiochemical properties of a relatively short AMP. Moreover, most fluorescent labels are placed on either the N- or C-terminus of the peptide, where they could further alter the net peptide charge by changing the ionization state of the terminal amine or carboxylic acid and/or by adding additional ionizable groups. One way to address these concerns is through the comparison of antimicrobial activity and bulk mechanistic assays, such as those for membrane permeabilization described above, for unlabeled and labeled peptides. However, investigations that consider increasingly subtle means for labeling will certainly be important in supporting future work to visualize the cellular localization of AMPs.

Recently, several thoughtful and inventive systems and assays have been developed whereby the use of multiple fluorescent molecules, both small probes and proteins, can provide insight about AMPs interactions with bacteria. The array of options in modern fluorescent probes and the availability of research-grade fluorescent microscopes afford the opportunity to address different facets of complex biological questions within a single experiment rather than in parallel or tandem. This includes the mechanism of membrane interaction, identifying intracellular targets for translocating peptides, and assessing broader physiological effects on proteins involved in critical cell functions.

Approaches that allow for timecourse measurements of peptide interactions with single bacterial cells

Using bulk bacterial cultures, it is difficult to obtain the time resolution necessary to image the first moments of an antimicrobial peptide interacting with bacteria. Even with the advantages provided by the incredible growth of fluorescence microscopy techniques and fluorophores, conventionally, samples of bacteria, AMPs and/or fluorescent labels have to be prepared prior to placement on the microscope for imaging. Depending on the location of laboratory space and imaging equipment, there may be a small, but potentially critical, delay between initial sample preparation and accessing the microscope. In addition, it takes the operator time to focus the sample and locate suitable cells to image. This means that imaging data is often captured some minutes after cells are first exposed to an antimicrobial peptide. Critical mechanistic information may be overlooked without the inclusion of data from the first seconds (subminute resolution) of bacterial and AMP interaction [61,67]. Moreover, the process to acquire images of enough individual cells for appropriate statistical analyses is often not trivial, and changes in membrane localization may occur as AMPs progress through different stages of their mechanism(s) of action.

Veldhiuzen and coworkers used the approach of immobilizing E. coli 506 in agarose and then placing the bacteria + agarose pad in a commercially available chamber compatible with an inverted confocal microscope (Attofluor Cell Chamber, Thermo Fisher Scientific, MA, USA). Fluorescently labeled peptide (FITC-CATH-2) and propidium iodide were then added, allowing for time-lapse imaging of peptide localization and cell envelope breach in live cells [68]. They demonstrated that FITC-CATH-2 reaches the cytoplasm within 1 min of exposure, followed by propidium iodide at around 90 s. This suggests a complex interaction with the cells, where the peptide translocates before the membrane is breached.

The Taheri-Araghi lab has also utilized agarose pad methods [69,70] to immobilize single bacterial cells to allow for their imaging over time after exposure to AMPs. In a recent paper, they used this approach to monitor the association between the slowed growth of a bacteria cell and a subsequent increase of AMP associating with that cell [71]. The ability of the researchers to collect over 380 trajectories of single cells allowed for them to develop mathematical models for peptide adsorption that could be compared with bulk culture measurements to provide an explanation for how minimum inhibitory concentrations (MIC) are impacted by bacterial density.

Building upon the E. coli expression system that produces periplasmic GFP, the Weisshaar lab has elegantly addressed time resolution issues through the use of customized microfluidic flow chambers [60,63]. While their microfluidic chamber systems have increased in sophistication since their inception, the general approach is to immobilize a population of live bacteria within a chamber where the medium containing AMPs and/or fluorescent probes can be changed while acquiring sets of images in both phase contrast and through multiple fluorescence channels on the order of every 2–20 s using a wide-field epifluorescence microscope. This effectively allows the monitoring of single cells before, during, and after the application of a peptide. Moreover, the simultaneous imaging of a population of cells allows for single-cell measurements during data analysis, which provides information on cell-to-cell heterogeneity within the imaged population.

Sochacki et al. first employed a microfluidic chamber to investigate the interactions of LL-37 with E. coli [63]. These studies examined the events as LL-37 moves through the Gram-negative cell envelope, with particular focus on how LL-37 affects the outer membrane and the cytoplasmic membrane. The microfluidic chamber methodology allowed them to document cell-to-cell differences in the time course of attack, potentially based on cell cycle, as growth is halted in septating cells first. They demonstrated that growth halting occurs before the cytoplasmic membrane is compromised. Though the exact mechanism and time point of lethality remained unknown, they hypothesized that LL-37 translocates across the outer membrane and binds to peptidoglycan, interfering with cell wall biogenesis in the septal region.

Choi et al. utilized a microfluidic chamber to examine, with a 12 s image cycle, the oxidative stress and damaged caused by the AMPs CM15 [61] and LL-37 [62] on E. coli. Using phase contrast and fluorescence channels, each requiring 6 s, they monitored cell length, the localization of fluorescently labeled CM15 [61] or of GFP and SYTOX Green [62], and the induction of oxidative stress using measures of autofluorescence from oxidized flavins, CellROX Green, and resorufin produced by the Amplex Red assay. They determined that exposure of both AMPs induces a rapid increase in reactive oxygen species through an interaction with the aerobic electron transport chain.

Their additional fluorescence microscopy-based investigations on live E coli. in flow chambers have shed light on the complex behavior of melittin with bacterial membranes [65] and also employed the smaller (702 Da) JF646 ligand in a HaloTag assay in place of the large (27 kDa) periplasmic GFP expression as the reporter of outer membrane permeabilization along with an increased time resolution of 3 s per image set cycle [64]. The use of microfluidic chambers in these two studies provided an avenue to develop a unifying sequence of membrane permeabilization events in live E. coli following exposure to several membrane active peptides. The Weisshaar group's most recent work connects flow chamber-based approaches with super-resolution microscopy via single-molecule localization to provide insight into the cytoplasmic interactions and killing mechanism(s) of LL-37 once the cytoplasmic membrane has been breached in live E. coli. [9].

Other researchers have focused on the use of microfluidic systems with GUVs, taking advantage of the control of vesicle composition to investigate the role of lipid composition on AMP activity. To this end, the Cooper lab has developed a device that allows for both the production of GUVs and their subsequent exposure to peptides, allowing them to follow the effect of peptides on a single vesicle over time in a high-throughput manner [72]. In a recent study, they used this device to study the effects that two peptides, melittin and m2a (a magainin analog), had on lipid vesicles with different compositions [73]. These measurements, which involved the collection of dye-leakage datasets for over 1500 GUVs, allowed them to develop and evaluate quantitative models of peptide selectivity for bacterial membranes.

Future perspective & summary

Over the past few decades, research on AMPs has increasingly transitioned in focus from the discovery of naturally occurring bioactive peptides to the characterization of AMP structure-function relationships and mechanisms of action. Fluorescence microscopy has played an increasingly central role in these efforts to understand how AMPs exert their biological activity. The ability to simultaneously measure multiple aspects of AMP function – such as membrane integrity, the colocalization of peptide with specific lipids or intracellular targets, or the ability of AMPs to enter cells – in a single experiment gives researchers the latitude to address a wide range of mechanistic questions. Moreover, the ability to quantitate fluorescence allows for precise biophysical measurements and the development of quantitative models of peptide action when desirable. Despite the inherent limits of light microscopy, researchers have innovated approaches to allow for the spatial and temporal resolutions needed to address pressing questions of AMP function. Thus, fluorescence microscopy approaches will remain central in the efforts to design increasingly effective AMPs for clinical use.

Executive summary.

• Antimicrobial peptides (AMPs) represent a promising alternative to conventional therapeutics in the face of increasing prevalence of antibiotic-resistant strains.

• Studies of antimicrobial peptides increasingly focus on characterizing how AMPs kill bacteria in order to promote efforts to design more therapeutically effective agents.

Fluorescence microscopy applications to analyze AMP interactions with bacterial membranes and entry into cells

• Fluorescence microscopy has become a central method for considering AMP mechanisms on the single-cell level.

• Effectively visualizing localization of peptides in bacterial cells requires imagining ‘slices’ through bacteria.

• However, distinguishing between slices that include membrane versus those that only include regions inside the cell is often compromised by the optical resolution limits of light microscopy.

Modified cells and lipid vesicles can help address resolution challenges in imaging bacterial cells

• Bacterial spheroplasts and protoplasts are modified versions of bacterial cells that have increased size and spherical shape that allow them to be more effectively visualized than normal cells.

• Giant unilamellar vesicles (GUVs) also have more favorable size and shape than normal bacteria for visualization.

• Both spheroplasts and GUVs have been used in studies providing important insights into AMP function.

Using multiple fluorescent labels can allow for imaging sublocalization of peptides or simultaneously measuring different mechanistic aspects of peptides

• Using fluorescently labeled peptides along with lipid membrane dyes can allow for more quantitative and systematic analyses of peptide localization while decreasing bias in imaging.

• Researchers using unique combinations of small-molecule fluorophores and fluorescent proteins are able to simultaneously assess different aspects of AMP mechanisms, such as membrane permeabilization and the targeting of an intracellular target, in a single experiment.

Approaches that allow for timecourse measurements of peptide interactions with single bacterial cells

• It is very difficult to image the first moments of an antimicrobial peptide interacting with bacteria in bulk culture, making researchers unable to observe rapid effects of AMP function.

• Researchers have overcome these issues of temporal resolution by employing microfluidic chambers and other approaches to immobilize bacteria allowing for characterizations of AMP mechanisms.