Live-cell imaging of actin dynamics reveals mechanisms of stereocilia length regulation in the inner ear

The maintenance of sensory hair cell stereocilia is critical for lifelong hearing; however, mechanisms of structural homeostasis remain poorly understood. Conflicting models propose that stereocilia F-actin cores are either continually renewed every 24–48 h via a treadmill or are stable, exceptionally long-lived structures. Here to distinguish between these models, we perform an unbiased survey of stereocilia actin dynamics in more than 500 utricle hair cells. Live-imaging EGFP-β-actin or dendra2-β-actin reveal stable F-actin cores with turnover and elongation restricted to stereocilia tips. Fixed-cell microscopy of wild-type and mutant β-actin demonstrates that incorporation of actin monomers into filaments is required for localization to stereocilia tips. Multi-isotope imaging mass spectrometry and live imaging of single differentiating hair cells capture stereociliogenesis and explain uniform incorporation of 15N-labelled protein and EGFP-β-actin into nascent stereocilia. Collectively, our analyses support a model in which stereocilia actin cores are stable structures that incorporate new F-actin only at the distal tips.

H air cells of the inner ear transduce sound energy and head movement into afferent nerve signals that are transmitted to the brain. Hair cells owe their name to the staircaseshaped bundle of mechanosensory stereocilia (Fig. 1), which are actin-based structures that project from the apical surface into the potassium-rich endolymph of the cochlear duct and the vestibular labyrinth. These mechanosensitive cells are terminally differentiated in mammals and are not regenerated when they die 4,5 .
The stereocilia bundle develops from a patch of microvilli on the apical surface of a differentiating hair cell. Each microvillus undergoes a complex process of programmed elongation and thickening by increasing the length and number of polarized parallel actin filaments (F-actin) that form the paracrystalline core of each stereocilium 1,2,6 . The barbed (plus) ends of the actin filaments within each core terminate near the distal tip, which is presumed to be the site of monomer (G-actin) addition during stereocilia development 3 . Conversely, the pointed (minus) ends of the filaments terminate near the apical surface of the hair cell or perhaps extend into the rootlets that anchor stereocilia in the cuticular plate 3,7,8 . Several actin crosslinking proteins including espin, plastin/fimbrin, fascin-2 and TRIOBP contribute to the rigidity of the F-actin cores in stereocilia [8][9][10][11][12] . The staircase architecture ( Fig. 1) and stiffness of the stereocilia bundle are crucial for the exquisite sensitivity of hair cells to mechanical displacements.
In contrast to the dynamic nature of microvilli, which are continuously created and disassembled with a half-life of a few minutes 13 , mammalian stereocilia are proposed to last for the lifetime of a hair cell. Evidence from transgenic mice suggests that once irreparably damaged, individual stereocilia are resorbed by the hair cell and not replaced 14,15 . It follows that the stereocilia actin cores of post-mitotic hair cells must be precisely maintained throughout the life of the organism. However, the molecular mechanisms that preserve these crucial mechanosensitive organelles are largely uncharacterized.
One intriguing model to explain the longevity of stereocilia is that their F-actin cores are continuously renewed via a perpetual actin treadmill. This model requires that actin monomer removal (depolymerization) at the base occurs at exactly the same rate as actin monomer addition (polymerization) at the distal tip in order to strictly maintain the steady-state length of each stereocilium. In this rapid turnover (treadmill) model, complete turnover of the F-actin core occurs every 24-48 h [16][17][18] . Actin treadmills are well studied in vitro 19,20 , and in microvilli, lamellipodia and dendritic spines in vivo [21][22][23] . Although actin turnover in stereocilia has been mathematically modelled 24 , direct experimental measures of polymerization and depolymerization rates have not been reported. Retrograde flow of newly incorporated and recycled G-actin in stereocilia cores was inferred from individual static images of paraformaldehydefixed neonatal hair cells at various time points following viral or biolistic gene gun transfection (propulsion of gold particles coated with plasmid DNA) expressing either enhanced green fluorescent protein (EGFP)-b-actin or EGFP-espin [16][17][18] .  In contrast to the rapid turnover model, the interpretation of data from other experiments indicates that stereocilia cores undergo slow protein turnover, on the order of weeks to months; a concept consistent with stereocilia cores containing exceptionally long-lived proteins (ELLPs) 25 . This model is supported by experiments in adult mice using multi-isotope imaging mass spectrometry (MIMS, measuring 15 N-leucine incorporation) and conditional mouse mutants in which Actb, encoding b-actin, expression is controlled by tamoxifen-inducible cre recombinase 26 . These data demonstrated a 'hotspot' of protein turnover that was restricted exclusively to the distal tip compartment in stereocilia 26 . Similar to the experimental evidence interpreted as supporting a rapid turnover model, the slow turnover model was inferred from a collection of individual static images of fixed hair cells from different animals at various time points.
In this report, we use continuous live-cell imaging in single hair cells to resolve these conflicting models. In addition, we provide MIMS data and high-resolution confocal images of over 400 stereocilia bundles from fixed hair cells transfected with EGFP-bactin or with mutant forms of EGFP-b-actin that are unable to polymerize into filaments. Our data show that only actin in the distal tips of stereocilia is rapidly renewed and polymerization dependent, while stereocilia cores are stable structures.

Results
Visualization of EGFP-b-actin in living hair cells. We hypothesized that live-cell imaging of hair cells biolistically transfected (that is, by gene gun) with EGFP-b-actin would enable us to collect longitudinal data from individual cells and would therefore reveal actin dynamics that cannot be observed using fixed samples. Given that utricular hair cells have longer stereocilia bundles, a higher transfection efficiency by gene gun, and were used for previous studies [16][17][18]26 we chose to image utricle explant cultures. In three independent experiments, spinning-disk confocal microscopy was used to record actin dynamics in 14 transfected hair cells that displayed EGFP-b-actin localization at the tips of stereocilia (Supplementary Movie 1). We reasoned that cells exhibiting tip localization of EGFP-b-actin would offer the maximum potential to observe rapid turnover of the F-actin cores postulated to occur in the treadmill model. In the majority of stereocilia after a minimum of 62 h of continuous live imaging, we did not detect movement of EGFP-b-actin from the tips towards the bases, as would be expected if a treadmill mechanism existed. Occasionally, a single stereocilium within a bundle displayed activity consistent with an F-actin treadmill; however, these unusual stereocilia were rare (Supplementary Movie 1). Instead, we consistently captured two distinct phenomena: (1) steady-state localization of EGFP-bactin in the distal stereocilia tips; and (2) asynchronous elongation of stereocilia (Fig. 1). In these cells, steady-state localization of EGFP-b-actin at stereocilia tips continued for up to 48 h. Beyond 48 h, elongation of some stereocilia within the bundle was recorded ( Fig. 1a and Supplementary Movie 1).
Expression of EGFP-b-actin during our live-cell imaging experiments was under the control of a cytomegalovirus (CMV) promoter. We considered the possibility that overexpression of EGFP-b-actin resulted in a constant increase in fluorescence intensity at stereocilia tips, causing a corresponding increase in dynamic range that might mask lower levels of incorporation within the F-actin core of the stereocilium shaft. If newly synthesized EGFP-b-actin is integrated into the cores of stereocilia in a manner consistent with a treadmill, then one would expect to observe a declining gradient of EGFP fluorescence from the tips towards the bases of stereocilia. To address this possibility, we performed line scans of stereocilia from three-dimensional (3D)-volume reconstructed bundles ( Fig. 2; Supplementary Fig. 1). The fluorescence intensity in the cuticular plate and distal tip compartment increased over time; however, we did not detect a corresponding increase or gradient of EGFP-b-actin along the length of the stereocilium shaft. We did detect a marginal, but uniform increase in fluorescence intensity along the length of the stereocilium that may represent a steady-state level of actin peripheral to the F-actin core (Fig. 2b). These data suggest that any incorporation of EGFP-b-actin into the F-actin core from the tips consistent with an F-actin treadmill is below the limits of detection.
Live imaging captures stereociliogenesis. While imaging the mouse utricular hair cells described above, we also captured de novo formation of stereocilia on the apical surfaces of seven EGFP-b-actin-expressing, newly differentiating hair cells. As the time-lapse experiment proceeded, the nascent stereocilia on these seven cells matured from microvilli and fluoresced green uniformly along their entire length consistent with the expression of EGFP-b-actin from the initiation of bundle formation ( Fig. 1b and Supplementary Movies 2 and 3). To examine the extent to which stereociliogenesis occurs in the neonatal utricle, we performed a second set of three independent experiments in which we imaged every EGFP-b-actin-transfected cell in each utricle culture. The plasmid-coated gold particles used in gene gun transfection are 1 mm in diameter and are expected to periodically strike and damage stereocilia bundles (Fig. 3a). Therefore, hair cells in which stereocilia were fused, bent or missing (39/112) were excluded from the analyses ( Fig. 3b-d, Table 1, Supplementary Movies 4-6). While actin dynamics at the tips of stereocilia may be influenced by a functional MET channel 27 , the F-actin core is not known to be directly affected by the presence of tip links, and therefore splayed bundles were included in the analyses. We found that stereocilia bundles in these additional 73 EGFP-b-actin-transfected hair cells all demonstrated one of the following two types of actin dynamics: steady-state tip localization with or without asynchronous elongation, or stereociliogenesis. Therefore, out of a total of 73 cells over six independent experiments, EGFP-b-actin remained steady state at the tips of morphologically stable stereocilia with or without asynchronous elongation in 53% of cells (29/73). Stereociliogenesis was observed in 60% of cells (44/73; Table 1), consistent with data indicating that 50% of mouse utricular hair cells differentiate and develop hair bundles during the first two weeks of postnatal development 28,29 . All of these live-imaging movies are available for viewing (Supplementary Movies 1-6).
EGFP-b-actin in stereocilia of fixed hair cells. Given that our data obtained from live-cell imaging are in direct opposition to previously published static, fixed-cell images [16][17][18] , we also performed fixed-cell confocal imaging of EGFP-b-actinexpressing hair cells. While live-cell imaging has the advantage of obtaining longitudinal data from individual cells, laser excitation intensity must be delicately balanced to limit EGFPphotobleaching and the generation of radical oxygen species. Confocal microscopy of fixed cells allows for the use of greater laser excitation intensity and a phalloidin counterstain to visualize the F-actin cores of stereocilia, and can yield images with higher resolution because the mounting media is optically matched to the objective lens. Similar to data we obtained with live-cell imaging, 49% (50/103) of biolistically transfected stereocilia bundles showed evidence of damage and were excluded from further analysis. We observed tip localization of EGFP-b-actin in 100% (14/14) of undamaged transfected stereocilia bundles 4 h post transfection (Fig. 4a, Table 1), consistent with previously published data [16][17][18] . However, in contrast to previous reports, we found that cells expressing EGFP-b-actin imaged 24 h post transfection displayed variability in the localization of the fluorescent signal. Specifically, 55% of morphologically intact transfected cells (22/40) still had EGFP-b-actin enriched at the tips of the stereocilia bundle, with weak uniform labelling of stereocilia shafts, whereas the remaining 45% (18/40) of bundles had uniformly strong labelling of stereocilia (Fig. 4a, Table 1). These data are consistent with our live-imaging data and support our conclusions that EGFP-b-actin dynamics in hair cell stereocilia fall into two distinct categories: (1) enrichment at tips with or without asynchronous elongation; and (2)   the lengths of stereocilia could be explained by diffusion from the cell body into the stereocilia bundle, we transfected neonatal utricle cultures with EGFP-b-actin mutants (either p.G13R or p.G63D) that fold correctly but are unable to incorporate into filaments 30,31 . We reasoned that EGFP-b-actin produced in the cell body might diffuse freely into the space surrounding the paracrystalline actin core and into the tip compartment without incorporating into the F-actin core. We transfected hair cells with either wild-type bactin, mutant b-actin or an EGFP-only negative control and imaged hair cells fixed at 4 and 24 h post transfection. Unlike wildtype EGFP-b-actin, we did not observe accumulation of mutant EGFP-b-actin G13R or EGFP-b-actin G63D at stereocilia tips in the vast majority (124/130) of fixed hair cells at 4 and 24 h post transfection (Fig. 4c,d, Table 1). However, EGFP-b-actin G13R or EGFP-b-actin G63D was present in the stereocilia shaft at a similar level observed for wild-type EGFP-b-actin. In 7% (6/90) of hair cells transfected with EGFP-b-actin G63D , we did detect enrichment of EGFP fluorescence at the tips of stereocilia that we did not observe in EGFP-b-actin G13R -transfected hair cells. These data are consistent with published biochemical data demonstrating that EGFP-b-actin G13R is incapable of polymerizing into filaments in the presence of wild-type b-actin 31 , while b-actin G63D can form small filaments in the presence of a high concentration of wild-type b-actin 30 . These data demonstrate that incorporation of actin monomers into filaments is required for trafficking of EGFP-bactin to the tips of hair cell stereocilia, retention of EGFP-b-actin at the tips or both. Control experiments in which hair cells were transfected with EGFP only (that is, no actin) demonstrated that complete fluorescent labelling of stereocilia shafts can be achieved solely by diffusion of proteins from the cell body (Fig. 4b, Table 1). Thus, our data indicate that some macromolecules in the cell body are not restricted from diffusion into the stereocilia bundle compartment.
To further investigate the dynamics of b-actin at the tips of stereocilia, we biolistically transfected a plasmid encoding    32 . We used photoconversion to track the movement of a specific population of differentially labelled fluorescent actin. Locations within dendra2-b-actin-expressing hair cells were specifically targeted with a 405 nm laser to photoconvert dendra2b-actin from green (dendra2-b-actin GREEN ) to red (dendra2-bactin RED ). Optical sections through the entire hair cell were taken every 30 min for 24 h. The dynamics of photoconverted dendra2b-actin RED were analysed using two independent methods (Fig. 5). We used Volocity 3D image analysis software to generate volume-rendered 3D images. Measurements were made at 6-h intervals to determine (1) the distance from the tip of a stereocilium to its insertion point in the hair cell cuticular plate and (2) the distance from the lower boundary of the photoconverted population of actin to the insertion point in the cuticular plate (Fig. 5). Using these methods, we found that the position of the photoconverted population of dendra2-b-actin RED remained stable relative to the stereocilium insertion point, while the total length of the stereocilium increased through addition of non-photoconverted dendra2-b-actin GREEN distal to the photoconverted dendra2-b-actin RED (Fig. 5, Supplementary Movie 7). To avoid any observer bias in these direct measurements, we also utilized Volocity to generate an automated method of defining and quantifying the distance from the photoconverted dendra2b-actin RED to the edge of the cuticular plate, which fluoresced green with non-photoconverted dendra2-b-actin GREEN (Fig. 5d). In agreement with our direct measurements, the softwaregenerated automated measurements indicate that the position of the photoconverted dendra2-b-actin RED remains stable until the point at which the signal is below the limits of detection (usually after about 16 h of imaging). These data are consistent with our observations using live imaging of EGFP-b-actin demonstrating that stereocilia can elongate their F-actin cores, but do not treadmill.
Developing hair bundles are detected by MIMS. In addition to live imaging, a second method was used to capture stereociliogenesis in postnatal utricles in vitro. We examined the incorporation of newly synthesized protein into growing stereocilia using MIMS 26 , a method that reveals the presence of selected isotopes in thin sections of tissue. To label newly synthesized protein with 15 N, a rare but stable isotope of nitrogen, we transferred newborn mouse pups to foster mothers whose food contained 15 N-leucine. Pups continued to nurse from these 15 N foster mothers for 4 or 15 days before sacrifice, so that protein synthesized after birth would be 15 N-labelled. We used MIMS to reveal the localization of both postnatally synthesized protein containing 15 N and prenatally synthesized protein containing 14 N, as previously described 26 . The location of postnatally synthesized protein was indicated as the ratio of 15 N/ 14 N at every point in the image (Fig. 6). At postnatal day 4, most hair bundles contained relatively low levels of 15  with our observations using both EGFP-b-actin and photoconvertible dendra2-b-actin, are consistent with a model in which stereocilia F-actin cores, once formed, do not treadmill. Instead, stereocilia are exceptionally long-lived structures with newly synthesized F-actin restricted to the distal tips.

Discussion
The rapid and slow turnover models for the maintenance of hair cell stereociliar F-actin cores are seemingly irreconcilable [16][17][18]26 . ARTICLE several long-standing questions. Had we only observed images of different fixed samples at multiple post-transfection intervals, the rapid turnover stereocilia treadmill model would have been a plausible explanation for our data, given the well-documented role of actin treadmilling in other types of cellular protrusions, such as dendritic spines 21,22 . In fact, when we performed similar fixed-cell experiments (Fig. 4), two patterns of EGFP-b-actin localizations within stereocilia bundles at 4 and 24 h post transfection were observed, further underscoring the importance of following EGFP-b-actin dynamics in the same cell over time by live imaging. As a specific example, it is impossible to distinguish between elongation from the tips of stereocilia and a retrograde actin treadmill unless multiple time points are collected for the same hair cell bundle and 3D measurements are made. A few caveats need to be considered when interpreting our data. First, caution must be exercised when assigning meaning to the percentage of cells that show a particular EGFP-b-actin localization. Of the 539 cells that we imaged, 217 either died shortly after the initiation of the experiment or their stereocilia bundles were extensively damaged by the mechanical trauma of biolistic (gene gun) transfection. Successful biolistic transfection requires that the plasmid-coated 1 mm gold particles penetrate the cellular membrane without striking the prominent but fragile stereocilia bundle 33 . Thus, it is not surprising that a significant proportion of cells transfected with this technique are damaged. Moreover, the proportion of cells categorized as 'enriched at tips' versus 'fully labelled bundle' may reflect increased efficiency of transfecting a nascent hair cell compared with that of a mature hair cell. In mature hair cells, the dense, heavily crosslinked F-actin comprising the cuticular plate at the apical surface (Fig. 3a) is likely to hinder biolistic transfection, whereas the gold particle may more easily penetrate the developing hair cell through an immature cuticular plate. Finally, overexpression of EGFP-b-actin may itself damage a hair cell, as actin concentrations in cells are closely controlled 34 . In particular, it is unclear whether the observed asynchronous elongation of an individual stereocilium in a bundle is a normal occurrence, a response to damage from biolistic transfection or an artifact of overexpressing EGFP-b-actin in cultured hair cells. Intravital imaging of hair cells in live animals will be necessary to resolve these open questions.
Data obtained using non-polymerizable missense mutants EGFP-b-actin G13R or EGFP-b-actin G63D indicate that the actin population in the distal tip compartment of stereocilia is likely composed of F-actin. These data, along with evidence from a companion paper 35 that the tip compartment is renewed at a faster rate than the actin core, raise questions about the role(s) of this dynamic F-actin pool in this distinct subcellular compartment. In addition to its established functions in cell structure and motility 36,37 , F-actin also serves as a scaffold or barrier to restrict the trafficking of vesicles or molecules 38,39 . F-actin in the distal tip compartment of stereocilia may be an extension of the paracrystalline core or perhaps this distinct population of actin has an alternative role in regulating stereocilia tip dynamics in close proximity to the site of mechanoelectric transduction 27,40 .
A model of stereocilia dynamics in which a zone of rapid turnover is immediately adjacent to a long-lived F-actin core is not without precedence in other eukaryotic structures. ELLPs that elude turnover 25 are found in the nuclear pore complexes of non-dividing cells. Some nucleoporins have half-lives on the order of months to years, while structurally adjacent nucleoporins encoded by different genes are rapidly replaced 25 . Emerging evidence supports the idea that a protein's half-life is not necessarily uniform throughout the cell or even within the same subcellular compartment. Instead, a protein's stability likely depends on several factors, including posttranslational modifications, interacting partners and subcellular localization 41,42 . Our data are consistent with the view that actin and actin-binding proteins comprising stereocilia cores are ELLPs, though the modifications that allow this long-term persistence remain elusive.
The repair mechanisms of long-lived structures such as stereocilia cores remain to be discovered. For example, gaps in phalloidin-stained F-actin cytoskeletons of stereocilia are particularly abundant after noise trauma or in the absence of either bor g-actin 14,15 . Toyama and Hetzer 41 have suggested that unrepaired damaged ELLPs may contribute to aging processes. Furthermore, some of the homeostatic mechanisms that maintain hair cells and preserve stereocilia function include the secretion of protective molecules from supporting cells and active remodelling of protein complexes that comprise tip links [43][44][45] . A fuller understanding of homeostasis and repair mechanisms within each of the various stereocilia compartments during aging may have translational potential for preservation and restoration of healthy hearing.

Methods
Plasmids and biolistic transfection. For live-cell imaging experiments, human ACTB cDNA (NM_001101.3, identical amino acid sequence to mouse b-actin) cloned into the Green FP expression vector (BD Biosciences, San Jose, CA) was used to prepare gold particles for biolistic transfection as previously described 42 with minor modifications. Specifically, we used antibiotics in the culture media (see Organ culture and live imaging below) and reduced the BioRad Helios Gene Gun firing pressure to 100 p.s.i. in order to reduce damage caused by biolistic transfection. The p.G13R and p.G63D substitutions of b-actin were introduced into pEGFP-b-actin using site-directed mutagenesis (Stratagene, La Jolla, CA) with primers 5 0 -CGTCGTCGACAACCGCTCCGGCATGTG-3 0 , 5 0 -CACATGCCGG AGCGGTTGTCGACGACG-3 0 , 5 0 -CCCAGAGCAAGAGAGACATCCTCA CCCTGAA-3 0 and 5 0 -TTCAGGGTGAGGATGTCTCTCTTGCTCTGGG-3 0 , respectively (point mutation in bold). pDendra2-b-actin was cloned using InFusion (Clontech, Mountain View, CA) with primers 5 0 -GTGTACAAGACTCGAGC CACCATGGATGATGATATCG-3 0 and 5 0 -TAGATCCGGTGGATCCCTAGAAG CATTTGCGGT-3 0 to amplify the cDNA of ACTB and insert it between the XhoI and BamHI sites of pDendra2-C (Clontech, Mountain View, CA). The DNA sequences of all plasmid inserts were verified by Sanger sequencing.
Organ culture and live imaging. C57Bl6/J pups were killed by decapitation at postnatal day 2-5 in accordance with National Institutes of Health Institutional Animal Care and Use Committee-approved guidelines under animal study protocol #1263. Utricles were dissected in L-15 media (Life Technologies, Carlsbad, CA, USA) and cultured on 2 mg ml À 1 collagen matrices in DMEM/F-12 with 20 mM L-glutamine, 7% fetal bovine serum, and 1 unit per ml penicillin-G. One hour before biolistic transfection, growth media was replaced with antibiotic-free media. Transfections were done using a Helios gene gun and endotoxin-free plasmid DNA (NucleoBond Xtra EF, Macharey-Nagel). Twenty-four hours post transfection, cultures were transferred to phenol-red free media supplemented with penicillin-G and mounted under a platinum and nylon harp 46 at 37°C in a humidified 6% CO 2 atmosphere with the stereocilia facing the coverglass and objective. Cells transfected with pEGFP-b-actin were live-imaged on a Zeiss Cell Observer spinning disc microscope ( Â 63 plan-apochromat, NA 1.4) for 62-92 h. Z-stack volumes (138 Â 0.4 mm) were captured at 1-h intervals. For photoconversion experiments with pDendra2-b-actin, transfected cultured cells were imaged for 24 h and z-stack volumes were captured at 30 min intervals. Dendra2-b-actin was converted by exposing a defined region of interest to 405 nm laser light for 15 Â 1-ms bursts at 6% of the total laser power (UltraView, Perkin Elmer, Akron, OH, Laser Module 2). Post-acquisition image analyses were performed using ImageJ (http://imagej.nih.gov/ij/) and Volocity (Perkin Elmer, Akron, OH, USA). For automated analysis of dendra2-b-actin experiments (Fig. 5d), the red photoconverted populations of dendra2-b-actin and the green non-photoconverted dendra2-b-actin cuticular plate were identified mathematically. The distance from the centroid of the red population to the edge of the cuticular plate was measured for each time point.
Fixed-cell imaging. Mutant expression constructs pEGFP-b-actin G63D and pEGFP-b-actin G13R , the wild-type EGFP-b-actin control and the EGFP-only control were biolistically transfected into cultured utricles from neonatal mice (P2-P5) and fixed in 4% paraformaldehyde at 4 and 24 h post transfection for imaging 17,18,33 . After fixation, samples were washed in phosphate-buffered saline, counterstained with rhodamine-conjugated phalloidin and mounted in ProLong Antifade Gold (Life Technologies).
Multi-isotope imaging mass spectrometry. Mouse pups born to normal mothers were switched at birth to nurse from mothers who had been fed for a month or more with food containing 15 N-leucine (B1.25% relative to 14 N-leucine). Pups were killed after 4 or 15 days. Utricles were fixed overnight at 4°C in 4% formaldehyde þ 10% glutaraldehyde in cacodylate buffer, then processed for plastic embedding and thin sectioning.
We performed MIMS as described previously 47,48 . Briefly, an accelerated beam of Cs þ ions, focused to a 30-nm spot, was scanned across the surface of a thin section, sputtering molecules from the surface. Anionic molecules were accelerated back and analysed with a double sector mass spectrometer. 15 N was detected as the cyanide ion 12 C 15 N À at mass 27 and compared with 12 C 14 N À at mass 26 for each pixel in the scanned image. Incorporation was calculated from the natural abundance of 15 N and the relative abundance of 15 N in the food 26 . To show incorporation as a percentage in the image, we used a hue saturation intensity transformation in which the hue represents percentage and the intensity indicates reliability.