Register for Mechanobiology 2016!

Epidermal keratinocyte polarity and motility require Ca2+ influx through TRPV1
David M. Graham, Ling Huang, Kenneth R. Robinson, Mark A. Messerli


Ca2+ has long been known to play an important role in cellular polarity and guidance. We studied the role of Ca2+ signaling during random and directed cell migration to better understand whether Ca2+ directs cell motility from the leading edge and which ion channels are involved in this function by using primary zebrafish keratinocytes. Rapid line-scan and time-lapse imaging of intracellular Ca2+ (Ca2+i) during migration and automated image alignment enabled us to characterize and map the spatiotemporal changes in Ca2+i. We show that asymmetric distributions of lamellipodial Ca2+ sparks are encoded in frequency, not amplitude, and that they correlate with cellular rotation during migration. Directed migration during galvanotaxis increases the frequency of Ca2+ sparks over the entire lamellipod; however, these events do not give rise to asymmetric Ca2+i signals that correlate with turning. We demonstrate that Ca2+-permeable channels within these cells are mechanically activated and include several transient receptor potential family members, including TRPV1. Last, we demonstrate that cell motility and Ca2+i activity are affected by pharmacological agents that target TRPV1, indicating a novel role for this channel during cell migration.


The building and rebuilding of tissues and organs requires organization that begins at the single-cell level. Mechanisms that generate polarity at the single-cell level are common to polarity at the multicellular level (Nelson, 2003) and, therefore, help develop a foundation for engineering replacement tissues and organs. Whereas cytoskeletal elements provide morphological structure and polarity, a number of upstream signaling cues provide the information required to organize the cytoskeleton (Petrie et al., 2009). Ca2+ is known to play a role in directing cell polarity and guidance in highly polarized processes like tip growth (Messerli and Robinson, 2007), neurite extension (Zheng and Poo, 2007), and cell migration (Maroto and Hamill, 2007). Asymmetries in the distribution of the intracellular Ca2+ concentration [Ca2+]i provide spatial and temporal information, resulting in directed control of cellular extension and migration. Ca2+ influx in each of these processes has been linked to localized activation of mechanosensitive ion channels. In multiple cases, mechanically sensitive channels provide an additional degree of cellular polarization through functioning as chemical receptors, guiding neurite turning (Li et al., 2005; Wang and Poo, 2005) and lamellipodial extension (Wei et al., 2009). During tip growth and neurite extension, standing gradients of [Ca2+]i together with Ca2+ transients at the leading edge, are used to encode directional information (Messerli and Robinson, 2007; Zheng and Poo, 2007).

During cell migration, however, Ca2+ signals appear to play roles at both the trailing and leading edges. Cells extend the leading edge through protrusion of the lamellipodia in the direction of migration while simultaneously retracting the lagging edge (Lauffenburger and Horwitz, 1996). Relatively steady Ca2+ influx leads to cellular retraction at both the front (Tsai and Meyer, 2012) and the rear where it enables cells to lift off the substrate in order to move forward (Lee et al., 1999; Maroto and Hamill, 2007). Transient Ca2+ influx at the leading edge of migrating cells promotes lamellipodial extension during chemotaxis (Wei et al., 2009). Despite the close association of Ca2+ with these processes, cell migration still occurs in the absence of extracellular Ca2+ when an electrical polarizing signal is applied (Brown and Loew, 1994; Fang et al., 1998; Huang et al., 2009). These seemingly conflicting reports encourage further investigations into the importance of Ca2+ signals during migration and in the presence of external polarizing cues.

Therapeutic treatments that impose polarity on cells or direct outgrowth and migration would be of great significance in repairing highly organized tissues and organs. Applied electric fields have shown promise in this regard by directing neurite extension and cell migration in vitro (McCaig et al., 2005; Zhao, 2009). In the clinic, electric fields have been used in phase I clinical trials to promote spinal repair (Shapiro et al., 2005) and have been used successfully for decades in promoting the healing of chronic wounds (Gardner et al., 1999). However, the mechanisms by which cells sense and become polarized by electric fields remain elusive. It has been proposed that electric fields polarize cells by inducing Ca2+i gradients (Mycielska and Djamgoz, 2004; McCaig et al., 2005; Zhao, 2009). During galvanotaxis and galvanotropism, polarized responses are reduced by inhibitors of Ca2+ influx (McCaig et al., 2005); however, they remain unaffected when Ca2+ is removed from the culture medium or when Ca2+i is buffered by the intracellular Ca2+ chelator BAPTA (Brown and Loew, 1994; Fang et al., 1998; Palmer et al., 2000; Huang et al., 2009). Interestingly, weak DC electric fields and low-frequency AC electric fields (1 Hz) increase Ca2+i, in vitro (Cho et al., 1999; Huang et al., 2009; Dubé et al., 2012). Careful analysis of Ca2+i changes during galvanotaxis are necessary to resolve these controversies and may provide insight toward the molecular and physical mechanisms by which cells sense and respond to weak DC electric fields.


Ca2+-dependent Polarity and Migration

Cellular migration has long been understood to be dependent on Ca2+ signaling (Maroto and Hamill, 2007). We confirmed this association in primary zebrafish keratinocytes by exploring the effects of Ca2+ availability on cell polarity and motility. As migratory speed is a function of cellular locomotion speed and directional persistence (Lauffenburger and Horwitz, 1996), we have reported on both parameters. Cells in fish Ringer's solution that contained 1.8 mM Ca2+ displayed a bilaterally symmetric morphology with an average migration speed of 7.5±0.3 µm/minute. After replacement of extracellular Ca2+ with 1.8 mM free Mg2+ and EGTA (free Ca2+<1 nM) or buffering of cytosolic Ca2+ with the membrane-permeable form of the Ca2+ chelator 5,5′-difluoro BAPTA, migration speed and average net displacement were significantly diminished (Table 1). In the absence of extracellular Ca2+, cells became radially symmetric, giving rise to cell bodies encircled by the lamellipodium. Cells continued to migrate – albeit back and forth – over their point of origin. BAPTA loading of cells approximately doubled the cytosolic Ca2+ buffering capacity (Marks and Maxfield, 1990) and gave rise to cells with irregular morphology, resulting in tightly adhered fragments or fragments that randomly migrated in directions away from the freely moving cells. These countering events often caused cells to undergo severe stretching and sudden changes in trajectory. Cells treated with 0.5% DMSO behaved similarly to cells in normal fish Ringer's solution (Table 1). These results confirm an important role for trans-plasma-membrane Ca2+ influx in the maintenance of normal cell polarity and migration.

Table 1. Summary of cell motility analyses

Lamellipodial Ca2+ sparks

Motivated by the above findings, we mapped the spatial distribution of Ca2+i within migrating cells (Fig. 1A) and used an alignment algorithm to enable superimposition of the individual images to enable further analysis (Fig. 1B,C). Time-lapse imaging of the cytosolic Ca2+ indicator, Indo-1, showed that the cell body maintains a significantly higher [Ca2+]i than the lamellipodium (1.3±0.1 fold, (P<0.001, n = 5) (Fig. 1D), as is consistent with previous works (Maroto and Hamill, 2007). Whole-cell Ca2+ pulses were not detected in isolated, migrating cells, presumably because the long-term absence of serum during cell sheet dissociation, Indo-1 loading and experimentation. However, small regions of high [Ca2+], termed ‘Ca2+ sparks’, were identified both within the lamellipodium and along the outer periphery of the cell body (Fig. 1A). The mean standard deviation of [Ca2+] changes (Fig. 1E) and the coefficient of variation (the ratio of the standard deviation of the pixel intensities to the mean pixel intensities) (Fig. 1F) indicated substantially greater variance in the lamellipodia compared with the cell bodies.

Fig. 1.

Characterization of epidermal keratinocyte lamellipodial Ca2+ sparks. (A) The pseudocolored raw image shows the Indo-1 ratio from a time series of a migrating keratinocyte. The color bar represents the relative Indo-1 ratio. Scale bar: 10 µm. A greater density of Ca2+ sparks occurs in the lamellipod than the cell body. (B) Alignment algorithm map of cell area to determine centroid and relative cellular angle about the x-axis. (C) Superimposition of cell outlines after alignment from time series. (D) Projected mean of Indo-1 intensity values. (E) Projected standard deviation of Indo-1 intensity values for the same cell in panel D. (F) Illustration of the coefficient of variation of Ca2+ events in the lamellipodium and cell body. The dotted red line indicates the region of the cell from which the line scan was collected in (G). (G) Representative kymograph of Ca2+ sparks from 5-millisecond line-scans of a migrating cell. y-axis is time, x-axis is distance. Scale bar is 5 µm. (H) Kymograph of panel G presenting events 10% above the mean of the lamellipodial Indo-1 ratio. (I) Indo-1 ratio intensity plot over time from panel G (see respective orange and black arrows). Mean intensity = 104. 10% incremental threshold values above the mean are identified. (J) Histogram of Ca2+ spark duration from line-scans acquired at 3- and 5-millisecond resolution. (K) Mean duration of Ca2+ sparks. (L) Histogram of mean open time for inward conducting ion channels. (M) Histogram of Ca2+ spark area from line-scans acquired at 3- and 5-millisecond resolution. (N) Mean area of Ca2+ sparks.

High-speed line-scan imaging was used to characterize the temporal and spatial nature of the Ca2+ sparks during migration. Fig. 1G shows an example of consecutive line-scans with time along the y-axis and Fig. 1H shows the same image after imposing a threshold of 10% above the mean of the Indo-1 ratio from Fig. 1G. Examples of the relative amplitudes of Ca2+ sparks are shown, including the events 10–50% above the mean Indo-1 ratio (Fig. 1I). The distribution of the duration of Ca2+ sparks (Fig. 1J) was shown to have a mean of 8.7±5.6 and 10.5±5.4 milliseconds (n = 3 for each resolution) for 3- and 5-millisecond line-scans, respectively (Fig. 1K). For comparison, cell-attached patch clamp recording was used on migrating cells to determine a mean open time of 1.8 milliseconds (n = 5) for inwardly conducting ion channels on the cell body at resting potential (Fig. 1L). Whereas electrical recordings enabled capture of 1-millisecond events, both methods indicate a relatively low abundance of events lasting longer than a few tens of milliseconds. The Ca2+ spark area in migrating cells spanned a wide region (Fig. 1M) with mean areas of 2.5±0.8 and 2.2±1.3 µm2 (Fig. 1N) for scans collected at 3- and 5-millisecond resolution, respectively.

Lamellipodial Ca2+ sparks were further analyzed to determine their frequency and distribution during cell migration. Discrete, spatially localized increases in the Indo-1 ratio increased >10% above the mean of the lamellipodia for, on average, 21.7% of the time (n = 28, range 12.6–41.0%). Ca2+ sparks were then analyzed for their relationship to cellular turning. The results show that the ratio of the [Ca2+]i magnitude between the left and right halves of each lamellipodium is not significantly different from 1.0 (P>0.5) for any of the cells, indicating that the amplitude of the lamellipodial Ca2+ sparks is not associated with cellular turning (Fig. 2A). A ratio of 1 was determined for both analyses when the entire area of the lamellipod was used or when only Ca2+ sparks (10% above the mean lamellipodial Indo-1 ratio) were analyzed.

Fig. 2.

The frequency of lamellipodial Ca2+ sparks correlates with cell turning. (A) Ratio of the Ca2+ spark amplitude (left:right) of all migratory cells analyzed, showing no correlation with turning. (B) Ratio of the Ca2+ spark relative frequency (left:right) of all migratory cells analyzed showing a positive correlation with turning. (C–N) Representative images and corresponding relative rotation plots from populations of cells grouped into four different categories on the basis of their spatial distribution of Ca2+ sparks and cell turning. Color bar represents the relative frequency of events 10% above the mean lamellipodial Indo-1 ratio, ranging from 0% (black) to 100% (white). Images and plots demonstrate cells from non-turning (C–E), randomly turning without a significant asymmetric distribution of Ca2+ sparks (F–H), and left-turning (I–K) and right-turning (L–N) groups with an asymmetric distribution of Ca2+ sparks. Non-turning is defined as cells with a relative rotation <2 degrees/minute. Scale bar: 10 µm.

Conversely, a significant relationship (P<0.02) was observed between the relative frequency of the lamellipodial Ca2+ sparks and rotation rate (Fig. 2B). A greater frequency of Ca2+ sparks on the right correlated with cells turning clockwise and a greater frequency of Ca2+ sparks on the left correlated with turning counterclockwise. These results indicate that frequency and not amplitude of lamellipodial Ca2+ sparks correlate with cellular turning.

Further analysis of lamellipodial Ca2+ sparks during migration revealed three separate groups of cells. These cells showed (1) no turning (n = 5), (2) turning without a significant difference in the left:right ratio of Ca2+ sparks (n = 13) and, (3) turning with a significant left:right ratio of Ca2+ sparks (n = 10). Representative examples of the frequency of Ca2+ sparks and the relative degree of turning for the different groups are shown in Fig. 2C-N. A numerical summary of these groups is shown in Table 2. Cells that migrated without substantial turning did not have significant differences in the asymmetric frequency of Ca2+ sparks between the two halves of the lamellipodium (Fig. 2C-E). Turning did occur in the absence of a significant difference in Ca2+ spark distribution; however, the direction of turning was random, with seven cells turning clockwise and six turning counterclockwise (Fig. 2F-H). When a statistically significant difference in the left:right Ca2+ spark ratio was measured, cells turned to the side of the lamellipodium with the greatest frequency of Ca2+ sparks (Fig. 2I-N). The differences in the left:right ratio of Ca2+ sparks were consistently statistically significant, even when a more stringent threshold was used, i.e. 40% above the mean Indo-1 ratio. In Table 2, directionality of turning cells is presented as clockwise (right) or counterclockwise (left) for both groups (with and without a difference in Ca2+ spark frequency) to demonstrate that no significant difference in Ca2+ sparks existed for either direction during random turning. Time-accumulated Ca2+ sparks reveal spatial Ca2+ gradients in motile cells that correlate with turning.

Table 2. Summary of Ca2+ spark frequency and galvanotaxis analyses

An applied electric field increases lamellipodial Ca2+ sparks

Previously, we reported that weak DC electric fields increase [Ca2+]i in keratinocytes but do not generate an asymmetry in Ca2+ magnitude between the two sides of the lamellipodia of cells (Huang et al., 2009). Here, we tested whether DC electric fields act by generating an asymmetric distribution of the frequency of Ca2+ sparks. DC electric fields (100 mV/mm) were applied to migrating cells while imaging Ca2+i (Fig. 3A). The results indicate an increase in the occurrence of lamellipodial Ca2+ sparks for 90% of the cell population. In 50% of the cells, this caused a greater mean [Ca2+]i in the lamellipodia than the cell bodies (Fig. 3B) and a greater standard deviation (Fig. 3C). Overall this gave rise to a lamellipodial coefficient of variation of 1.15±0.05 (Fig. 3D), which is significantly greater than reported above in untreated cells (P<0.0001). Despite the greater levels of lamellipodial [Ca2+], the lamellipodia remained intact, did not retract and cells continued to migrate (Fig. 3E). Frequency analysis of the Ca2+ sparks indicates no consistent spatial increase correlating with cathodal turning (P>0.5). No statistically significant cathode:anode Ca2+ spark ratio during turning was found in six of ten cells, and two out of ten cells showed a significant cathode:anode ratio while turning toward the cathode and two out of ten cells showed a significant cathode:anode ratio in the opposite direction of turning. These data indicate that electric-field-directed migration does not produce an asymmetric distribution of Ca2+ sparks.

Fig. 3.

Weak DC electric fields (100 mV/mm) direct migration but do not generate an asymmetric lateral distribution of Ca2+ sparks. (A) Illustration of electric field experimental setup. Cells chosen for this study were migrating perpendicular to field lines prior to the applied electric field. (B) Projected mean of Indo-1 ratio for a representative cathode-migrating cell. Color bar represents relative pixel intensity. Scale bar: 10 µm. (C) Projected standard deviation of Indo-1 ratio for the same cell in panel B. (D) Illustration of the coefficient of variation of Ca2+ events in the lamellipodium for cells exposed to the electric field. (E) Relative rotation plot displaying the degree of rotation of cells over time in an applied electric field. Here, rotation towards the cathode is defined as ‘positive’ relative rotation.

Mechanosensitive Ca2+ influx

Various stimuli for plasma membrane Ca2+ channels were tested to identify the nature of the Ca2+ influx during migration and galvanotaxis. Gd3+, a blocker of voltage-gated and mechanosensitive ion channels (Yang and Sachs, 1989; Biagi and Enyeart, 1990), significantly reduced single-cell motility (Fig. 4A), similar to results from cells cultured in Ca2+-free medium (Table 1). Separation of spontaneous electrical versus mechanical activation was performed by monitoring Ca2+i using Fluo-4. Depolarization of the plasma membrane potential with 100 mM K+ produced no measureable change in [Ca2+]i (n = 3 cell sheets; data not shown) consistent with an earlier report (Lee et al., 1999). Mechanical activation of cellular Ca2+ channels was performed by imposing shearing stress through fluid flow across sheets of keratinocytes. Prior to flow, transient rises in whole-cell [Ca2+]i occurred for individual cells that remained connected in cell sheets. Upon application of a transient shearing flow, [Ca2+]i increased in cells (Fig. 4B) and decayed rapidly in the absence of flow. Further screening for mechanical activation of plasma membrane ion channels was tested using whole-cell patch clamp recording. Increased current influx occurred when gentle pressure was applied to the patch electrode (Fig. 4C). Reversal potential of these currents was near +5 mV. Taken together, these results demonstrate the presence of plasma membrane, mechanically gated, non-selective cation channels in epidermal keratinocytes. Plasma membrane nicotinic acetylcholine receptors (nAChRs) have been identified in mammalian keratinocytes (Kurzen et al., 2007) but do not seem to be present in zebrafish keratinocytes, because neither 100 µM or 1 mM acetylcholine gave rise to measurable changes in [Ca2+]i (data not shown).

Fig. 4.

Ca2+ influx occurs predominantly through mechanosensitive channels. (A) xy-trajectory plot of cells exposed to 10 µM Gd3+ for 1 hour. Scale bar: 200 µm. (B) Graph from transient shear stress applied to keratinocyte sheets indicates a transient increase in whole-cell [Ca2+]i during application of fluid flow (n = 3 cell sheets). (C) Whole-cell voltage clamp recording of a migrating keratinocyte in fish Ringer's solution. Three times gentle pressure was applied to the patch pipette, increasing the probability of mechanosensitive channels to be in the open state. Spontaneously activated channels are open briefly. (D) Tissue-specific gene expression analysis from adult zebrafish keratinocytes of two ion channel families containing mechanosensitive Ca2+-permeable channels. Genes found at multiple loci were amplified at consensus regions, with the exception of trpc4 and trpc5 that were analyzed separately. The chromosome number is given for genes that were found at multiple loci. Loading control is β-actin.

Two gene families of ion channels that have Ca2+ permeability and mechanosensitive properties are those of the transient receptor potential (TRP) and those of the epithelial sodium channel [ENaC, also known as sodium channel non-neuronal 1 (SCNN1)]. We screened for annotated members of these families in zebrafish using the Sanger Institute and GenBank genome databases. Tissue-specific gene expression analyses were performed on the 22 and 6 annotated members of the TRP and ENaC gene families, respectively. Gene-specific primers were generated and optimized using total RNA isolated from adult whole-body zebrafish samples (Fig. 4D). Several members of the TRP family, including trpa1, trpc6, trpm2, trpm4, trpm5, trpm7, trpp2, trpv1, trpv4 and trpv6, were consistently detected in adult keratinocytes. A single ENaC member, asic1.3, was weakly and variably detected with a reproducibility of two out of nine samples. These data are shown in the figure as they might be of importance. The recently identified mechanosensitive channel, Piezo1, was not included in our analysis but is thought to be expressed in zebrafish skin according to aberrant cell extrusion function within the epithelium of Piezo1-knockdown animals (Eisenhoffer et al., 2012). The keratinocyte expression profile reflects not only basal keratinocytes (tp63, Bakkers et al., 2002; Lee and Kimelman, 2002) but the outer epidermal layer (krt8, Martorana et al., 2001), non-keratinocyte epidermal cell lineages (grhl1, Janicke et al., 2010) which include epidermal ionocyte progenitors, mucus-secreting cells and, perhaps, peripheral neurons (asic1.3, Paukert et al., 2004). Melanophores appear to be largely absent from the expression profile (c-kit, Kelsh et al., 2000). The wider expression profile of genes within skin is probably because of a higher number of other cell types.

TRPV1 is necessary for motility

The role of mechanosensitive TRP channels detected in our expression analysis was characterized during keratinocyte migration. TRPM7 has been shown to couple Ca2+ influx to membrane tension in mammalian fibroblasts (Wei et al., 2009). The zebrafish tdob508 mutant cell line harbors a channel-inactive form of TRPM7 (Low et al., 2011) that is lethal early in development, with fish surviving only into embryonic and post-embryonic stages (Elizondo et al., 2010). We found that cell motility and polarity were unhindered in keratinocyte cultures from the tdob508 mutant line (from 3 days post fertilization, data not shown). TRPV4 has been shown to be activated by hypotonic cell swelling and pressure sensation (Suzuki et al., 2003; Vriens et al., 2004). The TRPV4 antagonist, RN-1734 (Vincent et al., 2009), had no inhibitory effect on cell migration at concentrations up to 100 µM (Table 1).

TRPV1 is a well-studied member of the TRP family and is best known for its activation by vanilloid compounds (e.g. capsaicin, resiniferatoxin), acid and heat. Mechanosensitive activation has also been associated with this channel (Birder et al., 2002; Rong et al., 2004; Jones et al., 2005; Daly et al., 2007). The functional role of TRPV1 during cell motility was explored by screening TRPV1-selective agonists and antagonists during cell migration. No significant change in cell motility was observed upon treatment of cells with 10 µM capsaicin, when compared to fish Ringer's solution or DMSO controls (Fig. 5A,B,E; Table 1). Treatment of cells with 1 µM resiniferatoxin resulted in a significant but subtle decrease in cell motility (Fig. 5F; Table 1). Conversely, treatment with TRPV1 antagonists, 10 µM capsazepine or 5 µM SB-366791, significantly and reversibly reduced cell motility, similar to motility of cells that were kept in Ca2+ free fish Ringer's solution or were loaded with BAPTA (Fig. 5C,D,G,H; Table 1). Migration analysis was started immediately after the addition of any drug in order to remain consistent with prior analyses even though noticeable effects on migration did not occur for 5–10 minutes.

Fig. 5.

Pharmacological treatment of keratinocytes with TRPV1 agonists and antagonists affect cell motility and Ca2+ dynamics. (A–H) xy-trajectory plots displaying the migratory paths of individual cells over 1 hour in (A) fish Ringer's solution (FR), (B) 0.5% DMSO, (C) 5 mM EGTA ([Ca2+]e <1 nM), (D) 100 µM BAPTA-AM, (E) 10 µM capsaicin (CAP), (F) 1 µM resiniferatoxin (RTX), (G) 10 µM capsazepine (CAPZ), and (H) 5 µM SB-366791 (SB). Scale bar: 200 µm. (I,J) Illustrations of the coefficient of variation of Ca2+ events in the lamellipodium of cells in FR and 10 µM capsaicin (CAP). (K) Graph showing the fold-change in whole-cell Ca2+ bursts from cells exposed to 10 µM capsaicin, 1 µM resiniferatoxin, 10 µM capsazepine, 5 µM SB-366791 and 0.5% DMSO. (L) Whole-cell patch clamp traces at −60 mV holding potential on single migrating keratinocytes exposed to 1 µM capsaicin and 30 nM resiniferatoxin. Agonists were loaded into the patch pipette and whole-cell recordings were taken immediately after ramp tests. (M) Graphs showing the population of cells with whole-cell Ca2+ increases upon addition of FR at pH 7.0 (top) and pH 5.0 (bottom) in percent.

In the presence of 10 µM capsaicin, a significant increase in lamellipodial Ca2+ sparks was monitored, giving rise to a coefficient of variation that was significantly greater (P<0.02) than in untreated controls (Fig. 5I,J). Whole-cell Ca2+ pulses were also stimulated in capsaicin-treated cells, although they occurred in less than 15% of the total number of images. In the presence of TRPV1 inhibitors, cells lost their defined lamellipodia, an effect that prevented Ca2+ spark analysis. As an alternative, we characterized the changes in whole-cell Ca2+-pulse activity, which is also due to Ca2+ influx through plasma membrane mechanosensitive channels (Lee et al., 1999; Huang et al., 2009). Spontaneous whole-cell Ca2+ pulses from cell sheets increased ∼7-fold and 14-fold in response to 10 µM capsaicin (P<0.02) and 1 µM resiniferatoxin (P<0.03), respectively (Fig. 5K). Treatment with 10 µM capsazepine (P<0.01) and 5 µM SB-366791 (P<0.03) resulted in an ∼7-fold and 4-fold decrease, respectively.

During whole-cell patch clamp recording with 1 µM capsaicin or 30 nM resiniferatoxin in the patch pipette, channel activity in the plasma membrane increased (Fig. 5L). In the absence of drugs, single-channel currents were fewer and very brief. The trace for capsaicin shows increasing activation of plasma membrane currents as capsaicin diffuses throughout the cell. Resiniferatoxin commonly caused large cellular currents immediately after whole-cell clamp was achieved (n = 7). The resiniferatoxin trace in Fig. 5L shows progression of ion channel activity as the drug diffuses through the cell. Capsaicin-activated currents were outwardly rectified (−16.9 pA at −60 mV, +169.0 pA at +60 mV, n = 3) – a common feature of TRPV1 channels. Taken together, chemical treatment of epithelial keratinocytes with known TRPV1 agonists and antagonists modified Ca2+i in a manner that is consistent with interaction of TRPV1; however, only the block of TRPV1 channels significantly reduced migration.

TRPV1 is also activated by acidic conditions. No change of [Ca2+]i was observed in cells treated by gently adding fish Ringer's solution pH 7.0, removing possible activation contributed by shear stress (Fig. 5M). Exposure to fish Ringer's solution pH 5.0 caused a robust yet transient increase in whole-cell [Ca2+] in all cells within the population. This acid-induced Ca2+ rise decayed within ∼2 minutes (n = 5). In addition, only transient Ca2+i increases occurred when medium at pH 7.0 was completely replaced with medium at pH 5.0 by using a flow system (n = 3, data not shown).

In zebrafish, TRPV1 has only been reported in sensory neurons during embryogenesis (Caron et al., 2008). According to multiple sequence alignments of the region spanning transmembrane regions 2–4, which contain residues that are critical for capsaicin sensitivity, zebrafish TRPV1 is more closely related to chicken and rabbit than to rat or human TRPV1 and may, therefore, possess less sensitivity to agonists (Fig. 6A). We did not observe behavioral responses in animals that were exposed to either capsaicin or resiniferatoxin starting at 24, 48 or 72 hours post fertilization (hpf). We report expression of trpv1 early in development, at the 4–8 cell stage, with increased expression during morphogenesis of the neural tube (14-somite) (Fig. 6B). Expression was detected within sensory neurons and the hindbrain (Fig. 6C). Lower expression was observed in the embryonic epithelium in animals 24 hours post fertilization (Fig. 6C arrows). In adult zebrafish, we report widespread tissue expression of trpv1 that differed from other TRPV members trpv4 and trpv6 (Fig. 6D). Furthermore, trpv1 expression was detected in the motile fraction of the adult epidermal epithelia from keratinocyte explants (Fig. 6E,F). These data indicate that trpv1 expression is not limited to sensory neurons but is also in the epidermal epithelium.

Fig. 6.

Zebrafish TRPV1 is expressed in the epidermal epithelium. (A) Sequence alignment of TRPV1 proteins from different species comparing the critical residues for vanilloid activation. Defined transmembrane regions (TM) are based on the predicted membrane topology of zebrafish TRPV1. Red arrows indicate residues important for capsaicin sensitivity and are listed according to the residue positions of rat TRPV1. (B) Gene expression of TRPV members during different stages of development. (C) In situ hybridization for trpv1 in an embryo 24 hours post fertilization (hpf). Arrows indicate staining in the epidermal epithelium. Scale bar: 250 µm. (D) Tissue-specific gene expression of TRPV members. (E,F) In situ hybridization of the motile fraction of an explant of adult epidermal epithelia for trpv1 using antisense- and sense RNA probes. Scale bar: 50 µm.


During cell migration, variations in Ca2+ signals over space, time and magnitude have made their role difficult to understand. We chose to study Ca2+ signaling by using one of the most rapidly migrating vertebrate cell types (zebrafish keratinocytes), which – at ambient temperature – maintain migratory speeds that are 5–10 times faster than those of mammalian keratinocytes at 37°C (Maiuri et al., 2012). Zebrafish keratinocytes maintain a high locomotion speed and a relatively persistent migratory direction in the presence of normal levels of extracellular [Ca2+], i.e. 1–2 mM. We demonstrate that both speed and persistence are Ca2+-dependent processes. Disruption of Ca2+i availability by either blocking Ca2+ influx or buffering Ca2+i results in a loss of bilateral symmetry and a decrease in persistent migration but does not inhibit lamellipodial extension or the ability of the cell to migrate. Lamellipodial extension occurs during low lamellipodial resting [Ca2+]i (Tsai and Meyer, 2012). We propose that the substantial decrease in net cell motility in the absence of extracellular Ca2+ is a consequence of the loss of cell polarity and the constant shifting of migratory direction in the absence of polarity. In support of this idea, DC electric fields rescue polarized migration of zebrafish keratinocytes in the absence of extracellular Ca2+. Under these conditions, cells maintained an average speed that is only 55% lower than that of cells in normal [Ca2+] (Huang et al., 2009). This speed is much higher than the speed reported here when cells migrate in the absence of both extracellular Ca2+ and a polarizing applied electric field.

In visualizing the distribution of free Ca2+i with Indo-1, we demonstrate a higher, relatively steady [Ca2+]i in the cell body and Ca2+ sparks in the lamellipodium. Higher [Ca2+]i in the cell body appears to be a conserved characteristic of motile cells (Maroto and Hamill, 2007). Few articles describe the positive signaling role of Ca2+ in the lamellipod. Ca2+ influx at the leading edge of macrophages generates cellular organization (Evans and Falke, 2007), whereas Ca2+ flickers promote lamellipod extension during chemotaxis (Wei et al., 2009). Untreated keratinocytes in the study by Wei et al. showed Ca2+ sparks in the lamellipodium but still maintained a significantly lower average [Ca2+]i than that found in the cell body. The lower background [Ca2+] in the lamellipod might provide a better means of detecting the brief, spatially restricted Ca2+ sparks. These brief signals might reflect the rapid speed and highly dynamic nature of fish keratinocyte migration and a requirement to react more quickly to external cues, compared to the slower migrating human fibroblasts that show Ca2+ flickers that last much longer (Wei et al., 2009). In the absence of extracellular Ca2+, there is no difference between [Ca2+]i in the lamellipod and the cell body, and no noticeable variation in cellular Ca2+ signals (Huang et al., 2009).

Keratinocytes migrate relatively straight or with a slow persistent turning in the absence of an asymmetric distribution of lamellipodial Ca2+ sparks. However, when there was a statistically significant asymmetry in the frequency of Ca2+ sparks between the two halves of the lamellipod, keratinocytes pivoted about the side with the greater number of sparks. We interpret the pivoting as a slowing of the migration of the side with the greater number of Ca2+ sparks. The amplitude of these events does not appear to play a causal role in cellular turning.

The varying levels of Ca2+ during cell migration are consistent with the Ca2+ model initially proposed for growth cone extension and retraction (Kater et al., 1988). In keeping with this earlier model and applying it to migrating cells, we think that long-duration Ca2+ signals, such as those found in the cell body or the long-term pulses found in the lamellipod (Lee et al., 1999; Tsai and Meyer, 2012), lead to membrane retraction. However, brief, spatially restricted sparks that give rise to a lower average Ca2+ signal, modulate the cytoskeleton and promote controlled, directed migration, as reported earlier (Wei et al., 2009). We suggest that the asymmetry in lamellipodial Ca2+ sparks that promotes turning is not sufficient to cause retraction but is sufficient to slow lamellipodial extension so that more rapid extension of the other half of the cell leads to cellular turning.

Do Ca2+ sparks direct migration or are they simply a result of migration? The fact that cells turn in the absence of Ca2+ spark asymmetries indicates that they are not just a result of turning. It also indicates that asymmetries in [Ca2+]i are not required for cellular turning. Work exploring chemotaxis has shown that Ca2+ sparks are required for directing migration (Wei et al., 2009). We chose to study galvanotaxis, a mechanism for polarizing cells that is proposed to work through similar pathways as those used during chemotaxis (Zhao, 2009). Epidermal keratinocytes polarize in the presence of weak DC electric fields, a process that for some time has been postulated to be linked to asymmetries in Ca2+i signaling (Messerli and Graham, 2011). During galvanotaxis, we observed an increase in lamellipodial [Ca2+]i in 90% of cells, as they polarized and migrated toward the cathode. The widespread increase in [Ca2+] indicates that a non-graded mechanism of Ca2+ influx is induced by the uniform electric field. Despite the comparisons that have been drawn between cellular polarization by chemotaxis and galvanotaxis with respect to Ca2+i signaling, we conclude that they are different.

Spontaneous Ca2+i sparks have been linked to spontaneous, localized increases in [Ca2+]i through L-type Ca2+ channels, ryanodine receptors and Ca2+-induced Ca2+ release in cardiac and smooth muscle (Kamishima and Quayle, 2003), through mechanical activation in skeletal muscle (Weisleder et al., 2012) and fibroblasts through TRPM7 (Wei et al., 2009). Because no measurable Ca2+ influx occurred during depolarizing conditions of the plasma membrane in zebrafish keratinocytes, Ca2+ influx probably occurs through channels with mechanosensitive and/or ligand-gated properties. We propose that electro-osmotic flow induced by the DC electric field (McLaughlin and Poo, 1981) is generating a shear stress within the cellular boundary layer that activates Ca2+-permeable mechanosensitive channels over the entire cell surface. In the absence of the DC electric field, we think that spontaneous Ca2+ influx is also owing to mechanosensitive channels. Under these conditions, the Ca2+ sparks provide an indication of the mechanical stress within the cell. Therefore, mechanical stress is not just displayed at the trailing edge of the cell but across the lamellipodium as well. These results support efforts to describe cell polarity control through membrane tension in the lamellipodium (Houk et al., 2012). The localized changes in membrane tension and Ca2+ influx might control Rac and Rho signaling (Jin et al., 2005; Evans and Falke, 2007; Tian et al., 2010), giving rise to cytoskeletal and morphological polarity.

Based on the mechanosensitive nature of Ca2+ influx in these cells, we screened two families of ion channels that are known to comprise mechanosensitive members. We identified ten TRP channels in the migratory fraction of epidermal explants. We found that inhibition of TRPV1, but not TRPV4 or TRPM7, impaired keratinocyte polarity and migration. TRPV1 expression has been defined mostly in sensory ganglia (dorsal root, trigeminal and nodose) (Caterina et al., 2000), although it has also been detected in epithelial cells of the bladder lumen (Birder et al., 2002), epidermal keratinocytes (Denda et al., 2001; Pecze et al., 2008), as well as other non-neuronal tissue (Southall et al., 2003; Li et al., 2007). We detected robust levels of trpv1 in trigeminal sensory neurons in 24-hour embryos, which is consistent with its well-documented role in neuronal sensory function. Low levels of trpv1 were detected in the heterogenous embryonic skin; however, trpv1 expression is prominent in adult epidermal explants. It is important to note that the somas of sensory neurons have a relatively large diameter, which undoubtedly influences signal detection through in situ hybridization. From our gene expression analysis, the fairly widespread detection of trpv1 in diverse adult tissue types might reflect a significant role for TRPV1 expression in non-neuronal cell types or might reflect contamination from tissue innervations, as TRPV1-expressing trigeminal afferents can project into the innermost epidermal layers (Cavanaugh et al., 2011). By studying the motile fraction of the epithelia that primarily consist of basal keratinocytes, we avoided such neuronal contribution to demonstrate trpv1 expression in the epidermal epithelium.

Reversible pharmacological perturbation of keratinocyte migration indicates a novel functional role for TRPV1. The significant and reversible decrease in cell motility and Ca2+i activity observed upon treatment with antagonists indicates that TRPV1-mediated Ca2+ influx is necessary for migration. Inhibition could not be caused by non-specific inhibition of voltage-gated Ca2+ channels or nAChRs, because neither membrane depolarization (through addition of K+) nor acetylcholine treatment gave rise to measureable Ca2+ influx. Cell motility was not changed in the presence of capsaicin and subtly decreased in the presence of resiniferatoxin, but lamellipodial Ca2+ sparks and whole-cell Ca2+ bursts significantly increased. Our finding is in contrast with work in the slower migrating HepG2 cell line, in which treatment with capsaicin doubled migration speed (Waning et al., 2007).

Amino acid substitutions in rat TRPV1 have shown that mutation Y511A resulted in a near-complete loss of sensitivity, whereas mutation S512T resulted in 2-fold decrease in sensitivity (Jordt and Julius, 2002). Furthermore, mutation M547L resulted in no significant change, whereas T550I resulted in an ∼10-fold loss in sensitivity with 1 µM capsaicin; however, 10 µM capsaicin evoked currents similar to those in wild-type TRPV1 (Gavva et al., 2004). Using the residue positions of wild type and mutant rat TRPV1 for reference, zebrafish TRPV1 contains residues with no loss (Y511), 2-fold loss (T512), no loss (L547) and no loss at 10 µM capsaicin (I550). According to primary sequence data alone zebrafish TRPV1 appears capsaicin-sensitive, with reduced sensitivity. It is important to note that capsaicin sensitivity may also be conferred through other surrounding residues that have not yet been identified. The whole-cell patch clamp recording and the Ca2+-imaging data reported here, show channel activation with capsaicin and resiniferatoxin. Possible reasons for the lack of behavioral responses to TRPV1 agonists in zebrafish at 24, 48 or 72 hpf may include low TRPV1 protein expression during early development, inefficient permeability through the epithelial mucosal boundary, and/or possible mechanisms that regulate TRPV1 function in different cell types (e.g. covalent modifications, cofactor binding, splice variants and isoforms). Channel characterization within the endogenous cell environment and gene expression studies would provide additional insight into the functional nature of this channel in zebrafish.

Whereas TRPV1 agonists do not appear to be cytotoxic to human keratinocytes over brief periods (Pecze et al., 2008), reduced cellular growth and increased apoptosis occur in epidermal cells of hair follicles during organ culture after a 5-day exposure to 10 µM capsaicin (Bodó et al., 2005). In zebrafish keratinocytes, TRPV1 agonists induced increases in [Ca2+]i that are intermittent and not chronic, and that do not result in cytotoxic effects for up to 2 hours of exposure. Following activation with capsaicin, TRPV1 is desensitized through Ca2+ influx (Caterina et al., 1997), which might prevent further Ca2+ influx in these cells. In neurons, however, where TRPV1 agonists can be cytotoxic, additional activation of voltage-gated Ca2+ channels through TRPV1-mediated plasma membrane depolarization may exacerbate Ca2+ influx, leading to toxicity. Pharmacological inhibition of L-type Ca2+ channels attenuates capsaicin-induced neuronal toxicity (Shirakawa et al., 2008). The absence of voltage-gated Ca2+ channels in fish keratinocytes might be important for their survival against TRPV1-mediated toxicity.

Last, the mechanosensitive nature of TRPV1 was originally cast in doubt when the TRPV1−/− mice displayed an avoidance behavior in response to pricking of the hind paw (Caterina et al., 2000). Subsequent studies have shown that TRPV1−/− mice show altered mechanical hypersensitivity in colonic afferents (Jones et al., 2005; De Schepper et al., 2008), bladder afferents (Daly et al., 2007), urothelial cells (Birder et al., 2001; Birder et al., 2002), osmosensory neurons (Ciura et al., 2011) and muscle (Ro et al., 2009). Mechanical activation of TRPV1, through either direct or indirect mechanisms, might be involved in the mechanics of Ca2+ influx during the migration of epidermal keratinocytes.

Materials and Methods

Keratinocyte explant cultures

Scales from an anesthetized adult wild-type zebrafish (Danio rerio) were removed and plated as described previously (Huang et al., 2009). Cell sheets containing primarily migratory keratinocytes migrated off the scale overnight at 28°C. Cell sheets were dissociated into single cells upon treatment with Ca2+-free fish Ringer's solution (5 mM HEPES pH 7.0, 116 mM NaCl, 2.9 mM KCl, 2 mM MgCl2 and 5 mM EGTA) for 10 minutes at ambient temperature. Scales were manually removed from the dish after treatment and cells were allowed to recover in normal fish Ringer's solution (5 mM HEPES pH 7.0, 116 mM NaCl, 2.9 mM KCl, 1.8 mM CaCl2, 10 mM glucose) for 10 minutes. For single-cell motility experiments, cells were further dissociated with a 2-minute treatment in EDTA containing fish Ringer's solution and were incubated in normal fish Ringer's solution.

Embryonic keratinocytes from the tdob508-mutant line were obtained from anesthetized fish. Embryos were dissociated with mechanical perturbation in Ca2+-free medium to obtain single epidermal cells. All animal experiments were performed according to approved guidelines.

Cell motility analysis

Time-lapse images of migrating keratinocytes were collected from keratinocyte explant cultures under specified conditions for 1 hour at ambient temperature. Average speed was calculated as the excursion of the cell body over each 30-second time point for each individual cell. Average distance was calculated as the sum of the squares of the differences between two consecutive xy coordinates for each time point. Single cells chosen for analysis did not contact other cells or debris during the entirety of the experiment.

The persistence index was calculated as the ratio of the average net displacement divided by the product of the mean speed and the time of the experiment. The range of the persistence index is zero (no persistence) to one (migrating along a straight line away from its point of origin).

Manipulation of Ca2+

Treatments modulating cytosolic Ca2+ were performed with Ca2+-free fish Ringer's solution (extracellular Ca2+ buffering) and normal fish Ringer's solution containing 100 µM 5,5′-difluoro BAPTA-AM (Molecular Probes, Carlsbad, CA, USA) (intracellular Ca2+ buffering). Cells were loaded with 5,5′-difluoro BAPTA-AM for 1 hour at ambient temperature in normal fish Ringer's solution containing 0.5% BSA and were gently washed three times in normal fish Ringer's solution prior to imaging.

Ca2+ imaging

Changes in [Ca2+]i were measured using fluorescent Ca2+ indicators Indo-1 AM and Fluo-4 AM and (Molecular Probes, Carlsbad, CA, USA). Cells were incubated in fish Ringer's solution containing 0.1% BSA (Sigma-Aldrich), 0.001% pluronic F-127 (Molecular Probes), and either 10 µM Indo-1 AM or 5 µM Fluo-4 AM for 1 hour at ambient temperature in the dark. After incubation, cells were carefully washed three times with normal fish Ringer's solution, left to stand for 30 minutes and gently rinsed again before imaging.

Indo-1 microscopy was performed using two-photon excitation on Zeiss LSM platforms. Indo-1 was excited with a Coherent Chameleon II laser at 705 or 720 nm and the emitted light was collected with either internal detectors through bandpass filters 390–465 nm and 500–550 nm or GAsP detectors through bandpass filters 414/46 nm and 510/84 nm with a 458 nm dichroic mirror. Fluo-4 was excited with 488 nm and emitted light was collected through a 525/45 nm filter with an internal detector. Rapid line-scans were acquired at 3- and 5-millisecond temporal resolution whereas time-lapse images of Ca2+ during migration were collected every 5 seconds. Four consecutive line-scans were averaged for both types of image to reduce noise such that each line scan took ∼3 milliseconds to acquire. 128×128 pixel time-lapse images took less than 0.5 seconds to acquire.

Image processing

Indo-1 images were lowpass filtered, a ratio of the images was determined, (short wavelength:long wavelength), masked by the long Indo-1 wavelength and multiplied by 255 in ImageJ (Rasband, 1997-2012). Binary images of the Indo-1 ratio stack were used to calculate the cellular centroid and the relative angle of its long axis by calculating the moment about the x-axis using Matlab (Mathworks, Natick, MA, USA). These values were used to align and rotate images so that their long axis was parallel to the x-axis.

The left:right pixel intensity ratio (left half:right half) from each half of a cell was determined for each image in an image stack and the average ratio for the entire image stack was plotted as a function of cellular rotation. The relatively high resting ratio of Indo-1 in the cell body enabled identification and removal from analysis.

Image stacks were resampled using Matlab to identify the distribution of pixels that are more than 10% above threshold. The 10% value was chosen based on earlier reports of the Indo-1 ratio in cells and the relative intensity changes measured by other fluorescent Ca2+ indicators in migrating fish keratinocytes (Baker et al., 1994; Lee et al., 1999; Doyle et al., 2004; Huang et al., 2009). This value provided an objective criterion for analysis and enabled good separation between the background variation of the lamellipodial intensity and the Ca2+ sparks. Threshold was applied to each image and pixels above threshold were counted. The resultant image for each image stack reflects the spatial distribution of the number of times that the relative Indo-1 ratio was above threshold. The left:right pixel intensity ratio of the two halves of the cell enabled comparison of relative frequency of Ca2+ sparks. The ratios were plotted against cellular rotation where negative rotation corresponds to counterclockwise rotation and the positive rotation corresponds to clockwise rotation.


Whole-cell and cell-attached patch voltage recordings were performed on isolated, migratory keratinocytes in fish Ringer's solution at ambient temperature. Cells were dissociated with Ca2+-free fish Ringer's solution as described above and treated for 30 seconds with 0.05% Trypsin-EDTA (Gibco, Carlsbad, CA, USA) to improve high-resistance seals. Micropipettes (VWR Scientific, West Chester, PA) were pulled and fire polished to give tip resistances of 2–3 MΩ. Backfilling solution for whole-cell recordings consisted of (10 mM HEPES pH 7.2,150 mM KCl, 2 mM MgCl2, 5 mM EGTA), whereas cell-attached patch electrodes were backfilled with fish Ringer's solution. Voltage clamp recording was performed with an Axon Axopatch 200B amplifier (Molecular Devices, Sunnyvale, CA, USA) running with pClamp 9.2 software. Current records were acquired at 20 kHz with the lowpass input filter set at 5 kHz. TRPV1 agonists, capsaicin and resiniferatoxin, were mixed in with the backfilling solutions prior to obtaining >3 GΩ seals. In whole-cell clamp recording capsaicin was maintained at 1 µM and resiniferatoxin ranged from 30–300 nM. During cell-attached patch, final concentrations of 10 µM capsaicin and 1 µM resiniferatoxin were used in the patch pipette.

Gene expression analysis

Total RNA was isolated from either whole body, skin (devoid of scales and dissected from the lateral side of the fish) or keratinocyte explant cultures. Whole-body and tissue-specific RNA was isolated from adult wild-type fish essentially as described (Westerfield, 2000). All tissue and/or cell lysates were extracted in TRI reagent solution (Ambion, Carlsbad, CA, USA) as described by manufacturer. Total isolated RNA was treated with DNase I (New England Biolabs, Ipswich, MA, USA) for 30 minutes at 37°C and purified with the RNeasy MinElute Cleanup Kit (Qiagen, Valencia, CA, USA). RNA was converted to complementary DNA by using RT-PCR with M-MuLV reverse transcriptase (NEB) and subsequent PCR was performed on designated targets. A full list of primers can be found in supplementary material Table S1.


Pharmacological treatment with TRPV1 antagonists and agonists for cell motility and Ca2+ imaging were performed with 10 µM capsazepine (Tocris bioscience, Bristol, UK), 5 µM SB-366791 (Tocris bioscience), 10 µM synthetic capsaicin (N-vanillylnonanamide; Sigma-Aldrich, St Louis, MO, USA), and 1 µM resiniferatoxin (Tocris bioscience). Control treatment used was 0.5% (v/v) DMSO. Treatment of 24–72 hours zebrafish was performed with 100 µM capsaicin and 10 µM resiniferatoxin in E3 medium (1 mM HEPES pH 7.4, 5 mM NaCl, 0.17 KCl, 0.33 mM CaCl2 and 0.33 mM MgSO4) for ∼1 hour at ambient temperature.

Whole-mount and explant in situ hybridization

The trpv1 RNA probe was prepared from plasmid pCDNA3-TRPV1 by PCR amplification of the full-length TRPV1 encoding region (2460 bp) and in-vitro transcription with SP6 RNA polymerase (NEB). The RNA probe against trpv1 was labeled with DIG 11-UTP (Roche, Indianapolis, IN, USA) and detected with an anti-DIG antibody (alkaline phosphatase Fab fragment) (Roche) using NBT/BCIP (Roche), as described previously (Thisse et al., 2004). Both wild-type developing embryos and adult keratinocyte explant cultures were fixed with 4% paraformaldehyde in phosphate-buffered saline and processed as previously described (Thisse et al., 2004).


Student's t-test was used to determine statistical significance for the comparison between two groups. Regression analysis was used to determine significance of cellular rotation versus Ca2+ spark amplitude and versus Ca2+ spark frequency.


We thank Steve Correia and Nikohl Graham for their assistance with statistics, Alexander Schier's lab for providing the pCDNA3-TRPV1 construct, Rob Cornell and Greg Bonde for the tdob508 mutant line, Casey Kraft and Douglas Richardson for their help with advanced imaging instrumentation, and Robert Prendergast for informative discussions.


  • Author contributions

    D.M.G., L.H., K.R.R. and M.A.M. designed experiments; D.M.G., L.H. and M.A.M. performed experiments; D.M.G. and M.A.M. analyzed data; and D.M.G., K.R.R. and M.A.M. wrote the paper.

  • Funding

    This work was funded by NIH P30GM092374 to G.G. Borisy, the Eugene and Millicent Bell Fellowship Fund in Tissue Engineering and the Herman Research Award to M.A.M. Deposited in PMC for release after 12 months.

  • Supplementary material available online at

  • Accepted July 11, 2013.


View Abstract