## Introduction

The ocean is a complex mixture of dissolved ions and organic macromolecules with a pH and ionic strength that vary significantly with geographical location. Over the past 500 million years or so, sessile marine invertebrates have evolved diverse adhesion mechanisms to suit the varied assortment of materials that constitute their settlement substrata, including rocks of differing chemistry, plants and the surfaces of other animals. Performing the complex chemistry of adhesion in such a variable environment is not a trivial matter, yet adhesion to surfaces is routine for most sessile marine invertebrates and it is unusual to find a naturally occurring surface to which barnacles, for example, cannot attach. Indeed, both generalist barnacle species and those that are highly surface-specific have evolved mechanisms that facilitate tenacious adhesion, under particularly hostile conditions, to surfaces that differ significantly in terms of roughness, chemistry, elastic modulus, surface energy and charge1. In addition, organic materials released into the ocean, particularly to a surface, quickly become a target for bacteria and may be damaged by digestive enzymes and other extracellular polymeric substances. Since many barnacles settle readily onto biogenic surfaces and persist there for years, their adhesives must have evolved in a long-running ‘arms race’, both against bacteria and organisms that need to remain clear of epibionts2. The bioadhesive systems of barnacles and other sessile marine invertebrates have in fact become so effective that most living surfaces that resist colonization do so via toxic means3, grooming4 or by sloughing5.

It is imperative that surface-colonizing forms of marine invertebrates remain securely attached to a chosen substratum during metamorphosis and growth; dislodgement usually equates to death. To be effective, their adhesives must overcome the adverse physical conditions at hydrated interfaces as well as the presence of biofilm. In light of these challenges, the adhesive used by the barnacle cypris larva (or cyprid) during permanent attachment to a surface presents an interesting model for the development of synthetic adhesives with wide-ranging applications. The adhesives of marine organisms, and of barnacles in particular, are therefore of significant commercial interest for novel biomedical adhesives and implants, being already optimized for conditions akin to those in the human body. However, improved knowledge of their mechanisms of action will also inform the design of novel coating technologies to combat marine biofouling6. Ship hull fouling costs between $180 and$260 million per annum for the US Navy7 alone and contributes to a loss in cruising efficiency of up to 86% for a typical Navy ship8. The economic and environmental consequences of fouling, including higher fuel bills, translocation of invasive species and greenhouse gas emissions, are also becoming increasingly high profile as environmental restrictions on the use of antifouling biocides are tightened.

The results of studies aiming to elucidate the nature of the permanent adhesive of cyprids have been ambiguous and often contradictory. Although Knight-Jones and Crisp19 first presented evidence for quinone tanning, Hillman and Nace20 obtained none. Hillman and Nace, on the other hand, presented the first evidence for a small quantity of lipid in the adhesive, an observation made by no other investigators. All early studies agreed that cyprid permanent adhesive is predominantly composed of protein, and this is consistent with what is known from studies of other marine bioadhesives.

Walker’s original description of the adhesive plaque morphology16, based on transmission electron microscopy, noted three well-defined strata (layers 1, 2 and 3 from the outside in). With light microscopy, however, only two layers were visible—a low-density reticulum making up the majority of the plaque volume and a high-density external ‘skin’, both staining for protein17. Walker attributed the stratification in the original TEM micrographs to graded curing of the adhesive in the presence of an environmental catalyst; more recent observations using confocal laser-scanning microscopy (CLSM) have cast doubt upon that explanation, however18. Using CLSM, only two well-defined layers were visible: a central proteinaceous bulk and an external non-proteinaceous layer. Based upon these observations it is our view that the high-density external ‘skin’ observed by Walker under light microscopy was, indeed, the intermediate layer 2 in the original TEMs. We propose that layer 1 in Walker’s TEMs is the non-proteinaceous layer that was observed by Aldred et al.18 This layer was probably removed during histological specimen processing in the early studies, which involved the use of ethanol, xylene, acetone and propylene oxide16, leaving only layer 2 as the outer skin. Based upon these recent insights and the apparent removal of layer 1 by organic solvents, we hypothesized that the true outer phase of the adhesive plaque is lipidaceous in nature. If so, this would constitute the first description of a lipid being used as a functional marine bioadhesive.

To fully investigate this phase, the present study employs two-photon excited fluorescence microscopy (TPEFM), broadband coherent anti-Stokes Raman scattering (BCARS) microscopy, and CLSM, combined with a suite of fluorescent probes for lipids, lipopolysaccharides, proteins and phosphoproteins, to image cyprid adhesive plaques and excised cement glands. We further apply image analysis techniques, such as three-dimensional (3D) segmentation, polygonal mesh and isosurface generation to spatially delineate the two phases of the expressed adhesive.

Here we show that the barnacle cyprid adhesive is a bi-phasic system; one phase composed of phosphoproteins and the other composed of lipids, and these phases are contained in two different granules within the cyprid cement glands. During adhesion the lipidaceous phase is released first into which the proteinaceous phase is secreted. We hypothesize that the outer lipidaceous phase creates a conducive environment for the curing and crosslinking of the proteins and protects the proteins from excessive hydration, spreading and bacterial degradation

## Results

### Two-photon microscopy of the cyprid adhesive plaques

When adhesive plaques were stained for proteins, rendered images defined with great clarity, the high-density protein skin covering a low-density proteinaceous reticulum (Fig. 1c,d, Supplementary Figs 1 and 2 and Supplementary Movie 1). High-magnification imaging of single embedded antennules, labelled specifically for primary amines, highlighted a void between the bulk protein core and bacterial growth on the true outer layer of the adhesive plaque, indicating a non-proteinaceous phase between the protein bulk and the ambient seawater (Fig. 2a–f, Supplementary Movies 2 and 3). It is noteworthy that were it not for the visualization of bacterial growth on the true outer layer, this non-proteinaceous phase of the cyprid permanent adhesive would have been missed entirely.

Figure 2e,f and Supplementary Movie 3 show only the outer phase of the plaque and the bacteria accumulated on it. Simultaneous labelling with various fluorescent probes specific to phosphoproteins and lipids indicated clearly the presence of a lipidaceous outer phase (Fig. 2g–l, Supplementary Figs 3 and 4 and Supplementary Movies 4–7) and allowed detailed analysis of the respective phases of the adhesive plaque. The total volume of the lipid phase was considerably less than that of the protein phase, as evidenced by the difference in fluorescence intensities between lipid and protein-labelled adhesive plaques. The former formed only a thin outer layer around the proteinaceous core (Fig. 2g–i).

### BCARS imaging of the cyprid adhesive plaques

Label-free BCARS chemical imaging of the adhesive plaques also showed a marked difference between spectra from these lipidaceous and inner proteinaceous phases. From the BCARS analysis, it was clear that one unique phase encircled the other from the interface through to the top of the plaque (Fig. 2m–o and Supplementary Fig. 5). The thin outer layer possessed higher lipid content than other regions analysed, retaining a constantly higher ratio of CH2:CH3 at the 2,875:2,930 cm−1 bands, but with a substantial decrease from the phenylalanine and amide I/II protein-specific vibrational modes at 1,006, 1,668 and 1,560 cm−1, respectively21,22,23 (Fig. 2o). The ratio between the CH2 and protein-specific vibrational modes, or lipid:protein ratio, is indicative of local concentrations of the two species in biological systems24,25. The thin layer surrounding the bulk material was, therefore, unambiguously lipid-containing. In contrast, the central bulk of the adhesive contained a much higher concentration of protein-specific chemical moieties than that of even the surrounding tissue (for example, antennules and adhesive disc) of the cyprid.

### Functional imaging of the cyprid cement glands

The presence of a functional lipid component in the adhesive necessitated a re-examination of the adhesive storage apparatus, that is, the cyprid cement glands and the contents of their two cell types. Previous histological investigations failed to detect a significant lipid component in either the α- or β-granules owing, perhaps, to the use of harsh organic solvents in traditional histological preparations. In fact, a 2006 study26 questioned the existence of a second granule type. Not only is this assertion shown here to be unfounded but the images of toluidine blue-stained sections in the 2006 paper are themselves indicative of two granule types, one of which appeared to be lipid (cyan in colour). In the present study, under physiological conditions and without the use of harsh organic solvents, we present conclusive evidence of two granule types, stored separately in large columnar cells within the cement glands (Fig. 3a and Supplementary Fig. 2d). Our observations, however, depart somewhat from those of Walker16 in relation to the biochemistry of the adhesive constituents and the arrangement of the cell types. Whereas α-granules stained positively for protein, as described previously, and constitute the bulk of the cement gland volume (Fig. 3c,g,h, Supplementary Fig. 6, Supplementary Movie 8), β-granules contain negligible protein and stained positively for lipids (Fig. 3b,d–f, Supplementary Fig. 7 and Supplementary Movie 9). We have also shown that the α-granules are primarily composed of phosphoproteins. Walker’s arrangement of β-granule-containing cells at the dorsal and ventral extremes of the gland seems to be another artefact of physical sectioning of specimens. Instead, we observed cells containing β-granules along the entire length of the gland, but confined to the lower half closest to the median collecting duct (Fig. 3e, Supplementary Fig. 7 and Supplementary Movie 9). Although cement gland cells contain only one granule type, of either proteinaceous or lipidaceous composition, the proximity of the β-granules to the median collecting duct would support their being exocytosed first, followed by the α-granules. This hypothesis was also supported by the fact that during initial stages of adhesive release, the released adhesive did not fluoresce when labelled with protein-specific stains, while it showed strong fluorescence when labelled with lipid-specific stains, implying a sequential release of adhesive components from the cement glands where the lipid phase is released first. The natural process of exocytosis has never been observed in cyprids, although artificial stimulation using catecholamines has been achieved9,27 (Supplementary Movie 10).

## Discussion

Based upon our observations, we propose a novel mechanism for the permanent adhesion of the barnacle cypris larva in which a lipidaceous primer, stored in the β-granules, is released first. We hypothesize that this material displaces water from the interface creating an environment conducive to the curing of a subsequent proteinaceous component, released by the α-granules (Fig. 4a,b). This environment would not be anhydrous per se, and need not be, since hydrated coacervate-like bioadhesives are perfectly adapted for adhesion to hydrated interfaces14. Rather, we suggest that excessive and uncontrolled dissolution of the adhesive is prevented by a pre-existing outer layer of lipid. The solvation shell created by the long, non-polar, hydrocarbon chains in the lipids will reduce the entropy of the plaque at the lipid–water interface, excluding water and preventing excessive hydration of the protein phase. It was also shown previously18 that this outer phase exerts a controlling effect on the spreading of the inner phase, depending on the chemical nature of the substratum, and may therefore be fundamental to the regulation of adhesive spreading.

Our evidence suggests that the lipid phase is released first, followed by the proteinaceous second phase. While it is conceivable that the two components could be released in a gradient, beginning with a high concentration of the lipidaceous phase and ending with the proteinaceous phase, the clear demarcation between phases observed by Aldred et al.18 suggests that this is unlikely. Established theory, based upon the observations of Walley28, presents an alternative and seemingly inaccurate hypothesis whereby adhesive components are mixed, post-exocytosis, in a muscular sack (Fig. 1a) before release to the surface. Walley originally proposed this only as a hypothesis based upon the available evidence but the process was never observed in vivo. Mixing of the components, followed by release and subsequent phase-separation ex vivo is precluded by our observations early in the process of cement release, where only lipid was evident on the surface.

As the proteinaceous bulk of the adhesive is released into the lipidaceous phase, the former becomes concentrated at the interface between the two materials, resulting in a protein-dense interphase region. From our observations it is also apparent that, being rich in proteins, the proteinaceous phase would serve as a nutrient source for benthic bacteria and that the lipid phase may have a protective role, reducing bacterial biodegradation as the proteins cross link and cure. The result is a solid proteinaceous core that embeds the antennules and maintains the cohesive structure of the plaque (Fig. 4c). This phase is surrounded and possibly even adhered to the surface by the lipid-based primer. Lipid was often observed as faint fluorescence between the adhesive plaque and the surface (Fig. 4b), but with a signal insufficient to allow isosurface rendering.

Although lipid is the major component of the putative primer, the true composition of the material is likely lipoprotein and/or lipopolysaccharide (Supplementary Fig. 7i,j). As the outer phase is sufficiently robust to survive days to weeks in the marine environment, we do not exclude a degree of crosslinking of functional groups, nor do we discount the possibility of any proteins in the external phase serving in the initiation and maintenance of adhesion to surfaces. AFM was used to demonstrate both hardening of the outer phase over the course of an hour or so post-secretion29 and the time-limited susceptibility of the adhesive plaque to attack by protease enzymes30.

## Methods

### Cyprid culture

Barnacle larval culture was carried out at Duke University Marine Laboratory in Beaufort, North Carolina, USA. Balanus amphitrite Darwin (1,854) (=Amphibalanus amphitrite) nauplii (planktonic feeding stage) were obtained from field-collected adults and were incubated for 4 days at 27–28 °C. Larvae were raised on a diet of excess Skeletonema cosatum. Following 4 days of incubation, cyprids (settlement-stage larvae) were separated from the culture through a series of fine mesh filters and were isolated, cleaned and then shipped to Clemson University via overnight carrier at 6–8 °C. Larvae were maintained at 6–8 °C until use in experiments18,35.

### Sample preparation for staining cyprid cement plaques

Cyprids maintained at 6–8 °C were acclimated to room temperature (RT). Cyprids were then deposited on #1.5 borosilicate coverglass (Thermo-Fisher) in a 1-ml drop of artificial seawater (ASW) (33‰) and allowed to settle at 23 °C in a humidified incubator for 6–8 h (ref. 18). Permanently attached cyprids that had not metamorphosed into juvenile barnacles were subjected to the various staining procedures described below.

### Cyprid cement gland dissection

Cyprid cement glands were dissected as described previously35,36. Briefly, the dissections were carried out on #1.5 borosilicate coverglass (Thermo-Fisher) with the aid of a Nikon SMZ 800 stereo-microscope. Cyprids were relaxed by adding a 7.5% (w:v) MgCl2 drop-wise to Petri dishes containing cyprids in ASW. Once the cyprids were relaxed, the adductor muscle at the junction of the bivalve carapace was severed using acid-etched tungsten micro-needles (Fine Science Tools Inc., No. 10130-10, Foster City, CA, USA). Cement glands were then dissected out, separated and cleaned from associated tissue debris, brain, eyes and antennules. The cement glands were also subjected to staining procedures.

### Nile Red staining for lipids

Nile Red is a lipophilic stain that exhibits environment-sensitive fluorescence37. It is not fluorescent in aqueous media but is intensely fluorescent in lipid-rich environments38. A 1 mM stock solution of Nile Red (Sigma-Aldrich cat no. N-3013) was prepared in DMSO. From the stock, a working solution at 800 nM was prepared by dilution in phosphate-buffered saline (PBS). Cement glands and/or plaques on coverglass were placed in Petri dishes containing 4% formaldehyde solution freshly prepared from paraformaldehyde in PBS and allowed to fix for 30 min at RT. Samples were then washed twice with PBS for 5 min each on an orbital shaker to remove the fixative. Glands and plaques on coverglass were immersed in Nile Red staining solution for 15 min at RT, followed by two washes for 5 min each with PBS on the orbital shaker to remove excess dye. A second #1.5 coverglass was placed gently on top of the coverglass containing specimen and a drop of PBS. The two coverslips were slightly separated and temporarily sealed using silicone grease (General Electric). Two-photon excitation and emission wavelengths: ex755 nm, band-pass emission filter at 535/50 nm.

### Oil-Red-O staining for lipids

For Oil-Red-O staning39,40, cement glands or plaques on coverglass were fixed with formaldehyde as described above and immersed in Oil-Red-O (ORO; Sigma-Aldrich cat no. O0625) staining solution for 20 min, followed by washing twice with PBS for 5 min each on an orbital shaker. ORO staining solution (100 ml) was prepared by adding 40 ml of dH20 to 60 ml of 12 mM solution of ORO in 100% isopropanol. The ORO-staining solution was allowed to stabilize at RT for 20 min and then filtered through a 0.2-μm filter (Nalgene). Coverglass were prepared for imaging as described above. Two-photon excitation and emission wavelengths: ex755 nm, band-pass emission filter at 535/50 nm.

### Staining for lipopolysaccharides

Glands and/or adhesive plaques were subjected to lipopolysaccharide staining using Pro-Q Emerald stain (cat no. P20495, Life Technologies, Carlsbad, USA)41. Although the Pro-Q Emerald manufacturer’s protocol is designed for staining polyacrylamide gels, it was modified in this case to stain glands and adhesive plaques. Briefly, formaldehyde-fixed and PBS-washed glands and/or plaques were treated with the carbohydrate-oxidizing solution (3% acetic acid and periodic acid solution, provided with the kit) for 15 min, washed twice with PBS for 5 min each on an orbital shaker and then incubated with the Pro-Q Emerald staining solution for 30 min. This was followed by two PBS washes for 5 min each on an orbital shaker. Coverglass were prepared for imaging as described above. Two-photon excitation and emission wavelengths: ex800 nm, band-pass emission filter at 535/50 nm.

### Differential staining with lipid and protein dyes

Cement glands and/or adhesive plaques on coverglass were stained with a solution of the lipid membrane dye FM 1-43 FX (Life Technologies) prepared in DMSO in ASW at a final concentration of 5 mM for 10 min. Specimens were then washed twice with ASW for 5 min each on an orbital shaker and fixed with freshly prepared PBS buffered 4% formaldehyde for 30 min. Fixed specimens were washed with PBS twice for 5 min each on an orbital shaker and then incubated with a 1.4 mM solution of fluoresceinamine in PBS for 30 min. Fluoresceinamine binds to primary amines in proteins18. Specimens were washed twice with PBS and coverglass prepared for imaging as described above.

### Phosphoprotein staining

Although the manufacturer’s protocol for Pro-Q Diamond phosphoprotein stain (cat no. MPP33300, Life Technologies)42 is optimized for staining SDS-polyacrylamide gels, it was modified to stain cement glands and adhesive plaques. Briefly, cement glands and adhesive plaques were fixed with formaldehyde, as described previously. PBS-washed specimens were incubated with the Pro-Q Diamond staining solution for 60 min at RT while gently shaking on an orbital shaker. Following staining, the specimens were washed with PBS three times for 10 min each while on the orbital shaker. Coverglass were prepared for imaging as described above. Two-photon excitation and emission wavelengths: ex800 nm, band-pass emission filter at 610/75 nm.

### Coomassie blue staining of adhesive plaques

Formaldehyde-fixed and PBS-washed plaques were stained with 0.01% Coomassie Brilliant Blue R-250 (Sigma) in 50% methanol and 10% acetic acid for 1 h at RT. Following staining, plaques were de-stained with 40% methanol and 7% acetic acid for 10 min on an orbital shaker followed by a gentle rinse with PBS and 10 min PBS wash on an orbital shaker. Two-photon excitation and emission wavelengths: ex700 nm, band-pass emission filter at 535/50 nm.

### Noradrenaline-induced exocytosis of cement granules

Cyprid cement glands were dissected on coverglass as described above. The coverglass were then placed on the microscope stage and a 20 μM solution of noradrenaline, prepared in 0.2 μm filtered ASW, was added drop-wise to immerse the glands. An image series was recorded at multiple time points in differential interference contrast mode.

Before carrying out any of the staining methods described above, unstained specimens (adhesive plaques and cement glands) were checked for auto-fluorescence at excitation and emission wavelengths used in the experiments using both single-photon CLSM and two-photon exited fluorescence microscopy. Control samples, which were carried through the staining procedures, but without the addition of stains, were also imaged to check for non-specific fluorescence and auto-fluorescence. Neither the unstained nor the control samples exhibited auto-fluorescence in the emission wavelengths used for the assays.

### Single-photon CLSM

CLSM on cement glands and/or plaques was performed as described previously35,43. Briefly, specimens were imaged on a Nikon Eclipse TiE inverted microscope (Nikon, Melville, NY, USA), equipped with a C1Si spectral confocal scan head, following laser lines, 405, 457, 476, 488, 514, 561, 638 nm and appropriate band-pass filter sets for the excitation and emission wavelengths of fluorophores used. Imaging was carried out using Nikon CFI Plan Fluor × 20, 0.45 numerical aperture (NA) and Nikon CFI Plan Apo × 60, 1.2 NA objectives. The microscope was also equipped with a CoolSnap-HQ2 monochrome CCD camera (Photometrics, USA) and a Lambda 10-3 Motorized filter wheel (Sutter Instruments, Novato, California, USA)44. Multi-channel time lapse of noradrenaline-induced cement gland exocytosis was imaged by setting up a ‘time lapse’ routine in Nikon NIS Elements software controlling the microscope that acquired images automatically over a defined period of time.

### Two-photon excited fluorescence microscopy

TPEFM was performed by coupling a mode locked beam of a Ti:Saphire oscillator at 140 fs pulses at 80 MHz repetition rate (Chameleon Ultra-II, Coherent, Santa Clara, CA, USA) to a Movable Objective Microscope (Sutter Instruments, Novato, CA, USA)45,46. The laser intensity delivered to the microscope was controlled using a pockel cell, model 302RM driver and a model 350-80-LA-02 EO modulator (Conoptics, Danbury, CT, USA). The laser beam was raster-scanned in the microscope via a pair of galvanometric scanners (Cambridge Technologies, MA, USA) and projected onto a back aperture of a Nikon CFI Apo LWD × 25, 1.1 NA water immersion objective through a scan and tube lens. The four-channel non de-scanned detector set-up consisted of four GaAsP photomultiplier tubes H7422P40-MOD (Hamamatsu, Middlesex, NJ, USA) with appropriate band pass filters for excitation and emission wavelengths of the fluorophores used in the experiments. The microscope’s X,Y,Z movements, laser intensity modulation by the pockel cell and data acquisition were carried out using MScan software (Sutter Instruments).

### BCARS microscopy

Coherent anti-Stokes Raman scattering47 is used as an analytical tool and an imaging contrast mechanism in the realms of materials and biological sciences47,48,49,50. BCARS contributes to the suite of traditional CARS capabilities by providing improved chemical selectivity and sensitivity over a 3,000 cm−1 spectral range. The ability of BCARS to chemically image complex samples for obtaining a broad spectral signal at very short image acquisition times (<30 ms per pixel of blue-shifted light) and at high spatial resolution (<1 μm3 per spectral voxel) arises entirely from CARS being a nonlinear, third-order optical process where, ωanti-Stokes=ωProbe+ωPumpωStokes (refs 51, 52, 53, 54). As a result, BCARS has the ability to non-invasively measure localized Raman signals at high speeds without auto-fluorescence of biological samples and in 3D sample space free of any sample preparation and labelling.

The coherent Raman (BCARS) microscope is detailed in the schematic diagram in Supplementary Fig. 8. In detail, the output (70-fs pulses centred at 830 nm and a 80 MHz repetition rate) from a single Ti:Sapphire oscillator (MaiTai-DeepSee, Spectra Physics, Santa Clara, CA, USA) was split into two separate beams: one for the pump/probe transitions and other for the Stokes transitions. The bandwidth of one beam for the pump and Stokes transitions (narrowband pulse) was reduced by adjusting the slit width in a 4-f dispersionless filter to 8 cm−1 at full-width half-maximum, with the centre wavelength at 830 nm. The other beam was sent through a photonic crystal fiber (Crystal Fiber- Femtowhite 800, NKT Photonics, Morgnaville, NJ, USA) for supercontinuum generation to produce a spectrally broad (850–1,200 nm) pulse and then collimated using a parabolic mirror. The two beams were then collinearly combined before entering into the side port of an inverted microscope (Olympus IX71, Olympus, Allentown, PA, USA). The beam was focused onto the sample using a × 40, 0.95 NA objective lens (Olympus) for a lateral resolution spot size of ~0.5 μm. The average power at the focal point of the beam was ~14 mW. The scattered light was collected by a × 60, 0.8 NA objective lens (Olympus) and then passed through an 830 nm notch filter and an 810 nm short-pass filter to remove the excitation light. After filtering, the blue-shifted BCARS signal was introduced into a spectrograph (SP-2300, Acton with 600 lines/mm grating) and measured with a charged-coupled device (CCD, DU920-BR-DD, Andor, South Windsor, CT, USA). The exposure time on the CCD was 30 ms per sample spatial location over a spectral region of 800–3,400 cm−1 across 1,024 pixels of the CCD chip. Images were produced by scanning the sample laterally and axially in 0.5 μm increments. Image acquisition and initial analysis were performed using a custom-developed LabView (LabView 2011, National Instruments, Austin, TX, USA) programme. BCARS spectral data processing included cosmic ray spike removal, spectral phase retrieval using the time-domain Kramers–Kronig approach to extract the resonant Raman signal, and baseline detrending using a cubic-spline algorithm55,56.

### Data analysis and visualization for CLSM and TPEFM

Optical Z-sections (Z-series) collected using TPEFM were first converted to multipage TIFF (tagged image file format) files using Mview image analysis programme (Sutter Instruments). Multi-page TIFFs were then converted to Image Cytometry Standard file formats (.ics) using NIS Elements (Nikon). Image files in .ics formats from CLSM and TPEFM were analysed using NIS Elements and Imaris (Bitplane, Zurich, Switzerland) data analysis and visualization software. Two-dimensional images were exported and saved in TIFF format and assembled into a figure panel using PowerPoint (Microsoft). 3D volume renderings created in NIS Elements and Imaris were exported as .AVI files for Supplementary Movies.

3D polygonal mesh surface renderings were created from the confocal and two-photon data sets using Imaris. Briefly, Imaris allows segmentation of the data sets and use of algorithms, like Marching Cubes, to extract a polygonal mesh of an isosurface from voxels of the 3D data set. Adobe Photoshop CS5 was only used for creating Supplementary Fig. 2. For that figure, images were inverted using the ‘invert’ function in Photoshop and grayscale gradient mapping was assigned to the images using the ‘gradient map’ function. Brightness, contrast and gamma were then adjusted for the images and the figure panel was assembled using Microsoft PowerPoint.

Certain equipment, instruments or materials are identified in this paper in order to adequately specify the experimental details. Such identification does not imply recommendation by the National Institute of Standards and Technology (NIST) nor does it imply the materials are necessarily the best available for the purpose.