Dr

Ea

b

a

ARR2AA

KNCSO

1

(epCrorlR

UT

0d

Cell Calcium 45 (2009) 369–383

Contents lists available at ScienceDirect

Cell Calcium

journa l homepage: www.e lsev ier .com/ locate /ceca

istribution patterns of the Na+–Ca2+ exchanger and caveolin-3 in developingabbit cardiomyocytes

ric Lina,b,1, Vivian H.Y. Hunga,b,1, Haruyo Kashiharaa,b, Pauline Dana,b, Glen F. Tibbitsa,b,∗

Molecular Cardiac Physiology Group, Biomedical Physiology and Kinesiology, Simon Fraser University, Burnaby, B.C., Canada V5A 1S6Cardiovascular Sciences, Child and Family Research Institute, Vancouver, B.C., Canada V5Z 4H4

r t i c l e i n f o

rticle history:eceived 12 May 2008eceived in revised form1 November 2008ccepted 8 January 2009vailable online 27 February 2009

eywords:eonate myocardiumalcium handlingodium calcium exchangerntogeny

a b s t r a c t

In adult cardiac cells the established mechanism of excitation–contraction coupling is by calcium-inducedcalcium release (CICR) mediated by L-type Ca2+ channels. However, in neonate cardiomyocytes, a CICRmodality involving reverse mode Na+–Ca2+ exchanger (NCX) activity predominates. This has been hypoth-esized to be due, in part, to the high expression levels of NCX in the neonate heart which drop severalfold during ontogeny. Very little is known about the nature of NCX distribution within the cardiomyocyteand how this might change with development given the significant differences in gene expression. Weinvestigated the spatial arrangements of NCX in developing rabbit ventricular myocytes with traditionalas well as novel image processing and analysis techniques. Using image segmentation, colocalizationanalysis was conducted at the whole cell, compartmental (cell periphery and cell interior) and objectlevels. Because NCX has been suggested to colocalize with caveolin-3 (cav-3) and perhaps form a signal-ing unit within caveolae, the spatial relationship of NCX relative to cav-3 was also examined in detail.NCX and cav-3 objects were found to be isolated islands of lit voxels that are present after thresholding.These objects were categorized into non-colocalized (0%), lowly colocalized (<50%) and highly colocal-

ized (>50%) subpopulations in both the interior and peripheral compartments. Our results show that NCXand cav-3 are distributed on the peripheral membrane as discrete objects and are not highly colocalizedthroughout development. 3D distance analysis revealed that NCX and cav-3 objects are organized with alongitudinal and lateral periodicity of about 1 �m and that NCX and cav-3 cluster appear to be mutuallyexclusive on the cell periphery. We conclude that despite the very significant decrease in NCX expres-sion with maturation, qualitatively there were no differences in NCX surface distribution or in the spatial

.

relationship to caveolin 3

. Introduction

Excitation–contraction (E–C) coupling via L-type Ca2+ channelsCav1.2) mediating calcium-induced calcium release (CICR) is thestablished working model for mature ventricular myocytes. In thisrocess, membrane depolarization activates the voltage-sensitiveav1.2, allowing an influx of extracellular Ca2+ into the cytosol. Thiselatively small sarcolemmal Ca2+ influx results in a larger release

f Ca2+ from the sarcoplasmic reticulum (SR) through the ryanodineeceptor (RyR2). RyR-mediated release of SR Ca2+ raises the cytoso-ic Ca2+ concentration several-fold, thereby initiating contraction.elaxation occurs as cytosolic Ca2+ is removed, mainly by reuptake

∗ Corresponding author at: Molecular Cardiac Physiology Group, Simon Fraserniversity, 8888 University Drive, Burnaby, BC, Canada V5A 1S6.el.: +1 778 782 3658; fax: +1 778 782 3040.

E-mail address: tibbits@sfu.ca (G.F. Tibbits).1 These authors contributed equally to this paper.

143-4160/$ – see front matter © 2009 Elsevier Ltd. All rights reserved.oi:10.1016/j.ceca.2009.01.001

© 2009 Elsevier Ltd. All rights reserved.

into the SR via SR Ca2+ ATPase (SERCA2a) and extrusion (forwardmode) by the sodium calcium exchanger (NCX1.1) into the extra-cellular space [1]. Under specific conditions, NCX is also capable ofoperating in reverse mode, resulting in Ca2+ influx.

Although Ca2+ influx occurs predominantly through Cav1.2 [1]in adult cardiomyocytes, recent functional data from our labora-tory strongly suggests that in neonatal rabbit myocytes Ca2+ influxis primarily through reverse mode NCX activity that also results inCICR, termed NCX-mediated CICR (NCX-CICR) [2]. The contributionof NCX Ca2+ influx corresponds with developmental expression lev-els of NCX, which are high at birth and decrease several-fold duringmaturation [3–8]. Concurrently, L-type mediated CICR contributionincreases during development as Cav1.2 expression and couplingwith RyR2 increases [7,9–11].

Considering the reversal potential and the Ca2+ transport rate ofNCX, it is hypothesized that in the neonate cardiomyocytes, a highlocal density of NCX, Na+ channels, and RyR proteins are organizedin a restricted microdomain to allow for effective reverse modeNCX activity leading to NCX-CICR [2,12]. While previous studies

3 lcium

ogies(cs

tTclcemabi

idsmIma2tcar

2

2

lFaaifd

2

amaopit(5ittraat

70 E. Lin et al. / Cell Ca

n NCX distribution in the neonate myocytes suggest a homo-eneous surface distribution [13,14], recent immunofluorescencemages in rabbit ventricular myocytes showed a punctate periph-ral NCX distribution [15]. We proposed that in early developmentaltages, a microdomain containing this multi-protein functional unitcouplon) exists in rabbit ventricular myocytes. Several possibleandidates may be responsible for structuring this microdomainpace including caveolae.

Caveolae are narrowed-neck membrane invaginations (diame-ers 50–100 nm) with distinct lipid and protein constituents [16,17].he principal protein structural component of cardiac caveolae isaveolin-3. Other groups have reported biochemical and morpho-ogical data from adult cardiomyocytes suggesting association andolocalization between NCX and caveolin-3 (cav-3) [18,19], how-ver; more recent studies have challenged this notion [20]. Oneight hypothesize that a subset of NCX is colocalized with an

nchoring molecule such as cav-3 in the early developmental stages,ut such colocalization declines as ventricular myocytes mature

nto adulthood.In this study, we investigate in detail the distribution of NCX

n the cardiomyocyte in a developmental period in which [NCX]ecreases several fold. We also determine the spatial relation-hip between NCX and cav-3 during development using confocalicroscopy in conjunction with various image analysis techniques.

n all 5 age groups tested (3 day, 6 day, 10 day, 20 day and 56-day-oldyocytes), we found the majority of NCX and cav-3 to be weakly

ssociated at the cell periphery. In developmental stages (10 day,0 day and 56 day groups) only limited colocalization was found inhe cell interior. However, 3D image analysis of peripheral NCX andav-3 objects showed that both proteins appear to be distributed inhighly organized manner but tended to have a mutually exclusive

elationship that did not appear to change with age.

. Materials and methods

.1. Isolation of ventricular myocytes

Animals were cared for in accordance with the principles estab-ished by the Canadian Council on Animal Care (CCAC). The Simonraser University Animal Care Committee approved the use ofnimals and the experimental protocols used in this study inccordance with the CCAC regulations. Ventricular myocytes weresolated from the hearts of New Zealand White rabbits of either sexrom five different age groups (days post-partum): 3 day, 6 day, 10ay, 20 day, and 56 day as described previously [8].

.2. Antibodies

Primary antibodies used were monoclonal anti-caveolin-3 IgG1ntibody (610420, BD Biosciences, Mississauga, ON, Canada), andonoclonal anti-NCX IgM antibody (MA3-926, Affinity Biore-

gents, Golden, CO). These monoclonal antibodies were chosen inrder to avoid possible non-specific interactions of rabbit derivedolyclonal antibodies with the rabbit myocytes. Secondary antibod-

es were affinity purified, isotype specific goat polyclonal antibodieshat were either anti-mouse IgG1 conjugated to Alexa Fluor 488A21121, Invitrogen) or anti-mouse IgM conjugated to Alexa Fluor55 (A21426, Invitrogen). These secondary antibodies react specif-

cally with the Fc portion of the immunoglobulin heavy chain of

he appropriate isotype, thereby allowing for distinction betweenhe two mouse-derived monoclonal antibodies. To minimize cross-eactivity, these antibodies were also highly adsorbed against otherntibody isotypes. Control primary antibodies raised against anntigen absent in mammalian cells were obtained from DakoCy-omation, Denmark (mouse IgG1, X0931; mouse IgM, X0942).

45 (2009) 369–383

2.3. Labeling

The isolated cells were prepared as described previously [21]before they were adhered to poly-l-lysine (Sigma) coated cover-slips at an appropriate density. Antibody solutions were applieddirectly onto the coverslips. Cells were incubated in anti-caveolin-3 antibody diluted in antibody buffer (75 mM NaCl, 18 mM sodiumcitrate with 2% goat serum, 1% bovine serum albumin, 0.05% TritonX-100, 0.02% NaN3) at 125 ng/mL for 4 h at room temperature. Cellswere then washed for 3× 10 min in antibody wash solution (18 mMsodium citrate containing 0.05% Triton X-100) before incubatingin anti-mouse IgG1 conjugated to Alexa Fluor 488 (4 ng/mL) for1 h at room temperature. After 3× 10 min washes, cells were incu-bated with anti-NCX IgM antibody (250 ng/mL) at 4 ◦C overnight.After the overnight incubation, cells were washed with 3× 10 minwashes before the last 1-h incubation of anti-mouse IgM conju-gated to Alexa Fluor 555 (4 ng/mL) at room temperature. After 3×10 min washes in antibody buffer, cells were subjected to an addi-tional 5 min wash in PBS before the coverslips were mounted ontoglass slides in Slow Fade Gold anti-fade reagent (Invitrogen) andsealed with nail polish.

To determine the level of non-specific staining of the isotype spe-cific secondary antibodies in these cardiomyocytes, control primaryantibodies were used. Control cells were stained in the same way asdescribed above but instead of anti-caveolin-3 and anti-NCX anti-bodies, control primary antibodies of the same isotype were used atthe same concentrations as their respective experimental primaryantibodies (cav-3, IgG1: 125 ng/mL; NCX, IgM: 250 ng/mL). Controlimages were acquired under the same settings as the experimentalimages. Only very dim, diffuse staining was observed from theseimages. In addition, to determine the specificity of the secondaryantibodies used, single stain control experiments with mismatchedsecondary antibodies were performed (i.e. cav-3 + IgM; NCX + IgG1).The staining pattern was very dim and diffuse, similar to thatobserved from isotype matched control primary antibodies exper-iments.

2.4. Image acquisition

Images were acquired on a Zeiss LSM 5 Pascal (software ver-sion 2.8) laser scanning confocal microscope equipped with a Zeiss63X/1.4 Plan-Apochromat oil immersion objective. The 543 nm and488 nm excitation beams were supplied by helium–neon and argonlasers, respectively. Images were acquired with a voxel dimensionof 100 nm × 100 nm × 200 nm (axial) and the pinhole set to oneAiry disc. Sequential acquisition was used to minimize signal bleed-through.

2.5. Image processing and analysis

After image acquisition, images were deconvolved using a max-imum likelihood estimation (MLE) algorithm (Huygens Pro 2.4.1,Scientific Volume Imaging, Hilversum, The Netherlands) using anempirically determined point spread function (PSF). A cell out-line was created for each image and was further segmented intoindividual voxels layers as previously described [15]. This studyincorporated an anti-noise mask to refine the cell outline, thedetails on how these masks were created can be found in thedata supplement. This refined cell outline was then used to iter-atively determine a threshold value. The peripheral compartmentwas defined as the first 3 layers of the cell outline and the interior

compartment was defined as the volume enclosed by the peripheralcompartment.

For each cell, the peripheral compartment was used to deter-mine an appropriate threshold value. The peripheral compartmentwas subjected to multiple threshold values ranging from 0 to 10,000

E. Lin et al. / Cell Calcium 45 (2009) 369–383 371

F numr here ts mbers e obj

“mwdovt

sCctNbecpo

etwedioWctm

tdd

ig. 1. An iterative approach to image thresholding. In the peripheral volume, theanging from 0 to 10,000, in 100 point increments, were used. The peak value, wubsequent analysis. In this example, “2000” is the peak value with the maximal nuhown. Note the connectivity between objects in “0” and “1000” and the loss of som

intensity values” in 100 point increments (16-bit images with aaximum intensity value of 65,535). The number of discrete objectsas then determined at each threshold value. Here, an object wasefined as an island of voxels above the threshold value surroundedn all sides by sub-threshold values. For each label, the thresholdalue that resulted in the maximal number of objects was used ashe binary threshold for all subsequent analysis.

Colocalization events were defined as voxels that wereupra-threshold in analogous positions in the NCX and cav-3.olocalization for each label was defined as the number of colo-alized voxels divided by the total number of voxels containinghat particular label per compartment. For example, peripheralCX colocalization equals the number of colocalized voxels dividedy the number of NCX supra-threshold voxels also in the periph-ral compartment. Colocalization analyses were performed at theompartment level (i.e. whole cell compartment, peripheral com-artment and interior compartment) as well as at the individualbject level (object specific colocalization).

For object specific colocalization, individual objects were cat-gorized into colocalized and non-colocalized objects withinhe peripheral and interior compartments. Colocalized objectsere defined as objects containing one or more colocalization

vents (more than 0% colocalization). Non-colocalized objects wereefined as objects containing zero colocalization events (0% colocal-

zation). We then analyzed colocalized objects as individual regionsf interest and determined their individual colocalization indices.ithin these colocalized objects, we determined the overall colo-

alization index, which is equivalent to defining a region of interesthat excludes all non-colocalized objects within a given compart-

ent.To investigate the effects of development on the spatial distribu-

ions of NCX and cav-3 at the cellular level we calculated separationsistances connecting each object to its immediate neighbors. Threeistance measurements were used: (1) NCX to NCX, (2) cav-3 to

ber of objects was determined at multiple threshold values for each label. Valueshe maximal number of events occurred, was used as the binary threshold for allof events. Images of the periphery at five different thresholding intensities are alsoects at “3000” and “4000”.

cav-3 and (3) NCX to cav-3. Distance analysis was conducted forboth peripheral objects and internal objects directly beneath theperipheral compartment.

For this analysis, objects were reduced into single pixel co-ordinates by determining the center-of-mass of each object. Usingthe above mentioned cell outline the 3D co-ordinates of the center-of-mass were converted to a 2D equivalent array to reduce thecomplexity of the calculations. Additional details on the conversionalgorithm can be found in the data supplement. Once this conver-sion to 2D space is complete Delaney triangulation was used toestablish connectivity between each NCX and cav-3 cluster as canbe seen in Fig. 7. In analyses that included a directional componentdistances were calculated along a ±15◦ window along the specifieddirection. Only distances less than 3 �m in length were included.

Internal separation distances were determined as per theperipheral calculations except distances were measured 3 voxel lay-ers in from the surface. Because of the restricted space within thet-tubules, the separation distances between NCX objects and cav-3objects were not measured.

It should be noted that these images are collected using lightmicroscopy with the resolution limitations being well establishedand the conclusions derived in this and similar studies must bemade with this in mind.

2.6. Statistics

Two statistical tests were used to establish statistical sig-nificances of age related changes. The first test used was theKruskal–Wallis H-test, a non-parametric one-way analysis of vari-

ance, and was used to establish that at least one of the age groupswas significantly different from the rest. This was then followedby pair-wise comparisons using the Mann–Whitney U-test, a non-parametric test analogous to Student’s T-test. For the purposes ofthe statistical analysis the parameter N represents the number of

372 E. Lin et al. / Cell Calcium 45 (2009) 369–383

Fig. 2. Object size as a function of circumference. To test the influence of differential lateral and axial resolutions on apparent protein localization, object size (number ofvoxels) and object locations were determined along the cell circumference. Because cell circumference varies with age, object positions were described by their angularrelation to the center-line of the cell. The center-line of the cell was determined by taking the center-of-mass of the equivalent cross-section of the determined cell outline.Therefore, the positioning of each object was described by the angle of the line connecting the center-line of the cell outline and the object’s center-of-mass. Orientation ofthe cells is shown in (A). Median object sizes per angular location for NCX and cav-3 are shown in (B) and (C). Higher magnification is shown in (D) and (E) and indicatesconstant object sizes on the top and bottom aspects of the cell.

Fig. 3. cav-3 and NCX distributions in developing ventricular myocytes. Images of cav-3 (green) and NCX (red) labeled ventricular myocytes presented as (A) 800 nm thick XYsurface sections, (B) middle XY section, and (C) 800 nm thick cross-sections from 3 day, 6 day, 10 day, 20 day, and 56-day-old rabbits. Both surface and cross-section images aremaximum intensity projections. Scale bar is 5 �m. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of the article.)

E. Lin et al. / Cell Calcium 45 (2009) 369–383 373

F e celluc gmenc roups( ery re

c62

3

3

etaewomtpviicibt

aiist

ig. 4. Compartmental colocalization. Colocalization analyses were applied to threav-3. (A) Indicates colocalization analysis from the entire cell without any image seompartments, respectively, as defined in the methods. Only the 10, 20 and 56 days gD) Shows the number of NCX, cav-3 and colocalized voxels found on the cell periph

ells included in the analysis. For 3 day, N = 14, from 2 animals; forday, N = 36, from 3 animals; for 10 day, N = 22, from 3 animals; for0 day, N = 50, from 6 animals; for 56 day, N = 23, from 5 animals.

. Results

.1. Iterative thresholding technique

Thresholding is an integral parameter to all imaging studies thatssentially defines the minimal signal required to be included in tohe analysis. Visually, NCX and cav-3 appeared as punctate objects,n area of bright pixels surrounded by an area of dim pixels. In anffort to find the balance between excluding dim but real objectsith the need to limit erroneous connectivity between individual

bjects, we utilized an iterative thresholding technique that appliesultiple threshold values and calculates the number of objects on

he cell periphery at each threshold value. Fig. 1A is a representativelot of the number of independent objects as a function of thresholdalue. Shown in Fig. 1B are images of a cell at different threshold-ng values. As the threshold increased from 0, the number of eventsncreased as dim connective regions were excluded. Each profileontained a peak value in which a threshold value generated a max-mal number of events. Any further increases in threshold valuesegan to remove dim objects. The threshold value that generatedhe maximal number of events was used in this study.

Although it is unlikely that dim objects were excluded from the

nalysis, it is possible that object size is excessively reduced by theterative thresholding technique. The median object size is shownn Fig. 2, panels B–E, as a function of the polar angle in the cross-ectional plane. Notable is the non-homogenous object size alonghe circumference of the cell for NCX and cav-3. Here, the left and

lar compartments to investigate possibly differing relationships between NCX andtation. (B and C) Indicate the colocalization in the individual peripheral and interiorare shown in (C) since only these age groups had appreciable t-tubule development.lative to those found in the whole cell.

right aspects of the cell align to the 0◦ and 180◦ marks and the topand bottom aspects of the cell correspond to the +90◦ and −90◦

marks. The median object size along the top and bottom aspects ofthe cell are noticeably smaller than object sizes along the left andright aspects of the cell. This indicates that the iterative threshold-ing technique does not excessively reduce object size. However, theanisotropic object size suggests that the axial resolution may affectwhere colocalization events are found.

3.2. Visual comparisons

Five age groups were considered in this study and images werecollected from 3 day, 6 day, 10 day, 20 day, and 56 day myocytespost-partum. Fig. 3 shows surface XY sections (A), middle XY sec-tions (B) and cross-sectionals (C) of representative cells usingiteratively determined threshold values. From the cross-sectionalimages, it is apparent that as the cells develop both NCX andcav-3 labeling increased in the interior of the cell. Images fromyoung cells (3 and 6 day) contain only minimal interior label-ing. In the 10 day group interior labeling was seen but did notappear robust. The 20 day and 56 day groups showed increas-ingly robust NCX and cav-3 labeling in the cell interior. NCX andcav-3 labeling in the cell interior appeared as periodic inwardprotrusions from the cell surface along what may be t-tubulestructures. Although changes in interior labeling were easily seen,we were unable to visualize any large changes in the NCX and

cav-3 distribution on the surface of the cell. NCX and cav-3appeared punctate and evenly distributed along the cell surface.Visually no definite pattern in colocalization was seen on the cellsurface but noticeable colocalization was observed in the inte-rior.

374 E. Lin et al. / Cell Calcium 45 (2009) 369–383

Fig. 5. Peripheral object-specific colocalization. To investigate the colocalization specific to the peripheral compartment, object-specific colocalization analysis was applied.(A) Indicates the percentage of NCX and cav-3 objects on the cell periphery that contained one or more colocalized voxels. The remaining balance of objects contained nocolocalization events and therefore has 0% colocalization. (B) Indicates the level of colocalization found if the 0% colocalization events were excluded from the calculation. Thisdecreases the size of the denominator proportional to the number of voxels contained in the 0% colocalization objects and therefore increases the colocalization index. Only asmall increase was found in (B) when compared to (A) indicating that the removed objects have relatively little mass. (C and D) Show cumulated object-specific colocalizationi e pres( coloco percec

3

atwrncCdda

ndices from the 20 day group. The histogram bin-size has been set to 1 to show thE and F) Show the quantification of highly colocalization objects (greater than 50%f peripheral objects that contained more than 50% colocalization. (F) Indicates theolocalization as shown in (E).

.3. Global colocalization in various compartments

Visually the peripheral and interior relationships between NCXnd cav-3 appeared to differ. To determine the relationship betweenhese two proteins, the colocalization index for three compartmentsas calculated. Fig. 4 shows whole cell (A), peripheral (B) and inte-

ior colocalization (C) between NCX and cav-3. Neither whole cellor peripheral colocalization showed a clear age dependency but

olocalization in the cell interior appeared to increase with age.hanges between neighboring developmental stages up to the 20ay group were found to be statistically non-significant (d3 vs. d6,6 vs. d10, and d10 vs. d20 but not d20 vs. d56) for the whole cellnd peripheral colocalization indices. P-values for the individual

ence of small ratios, peaks that can be seen at 100% (1:1), 50% (1:2), 66% (1:3), etc.alization) by object count or by voxel count, respectively. (E) Shows the percentagentage of peripheral colocalization voxels contained in object with greater than 50%

tests for all graphed variables are reported in the appendices. Inthe interior, the changes in development between the d10 and d20groups were not significant for interior NCX colocalization but weresignificant for interior cav-3 colocalization. In all these global colo-calization parameters, the 56 day group was found to be statisticallydifferent from the other age groups.

Because changes in the interior colocalization did not seem toaffect the whole cell colocalization indices, we also quantified the

amount of colocalization events in the periphery of the cell. Fig. 4Dshows the relative number of NCX, cav-3 and colocalization eventson the periphery as a percentage of total events and all changesbetween neighboring developmental stages were found to be sta-tistically significant.

E. Lin et al. / Cell Calcium 45 (2009) 369–383 375

F was at izatioc up. (ER more

3

scit

wtarnpi

ig. 6. Interior object-specific colocalization. Object-specific colocalization analysishat contained one or more colocalized voxels. (B) Shows the object-specific colocalumulated object-specific colocalization analysis from the interior of the 20 day groeports the number of colocalized voxels accounted for in those objects containing

.4. Object specific colocalization in the peripheral compartments

The low colocalization percentages found above may be due to amall population of highly colocalized objects amidst a lowly colo-alized general population. This low or non-colocalized populationf much larger than the highly colocalized population may dilutehe calculations, reducing the derived colocalization index.

To test for a highly colocalized subpopulation, we identifiedhich NCX and cav-3 objects contained one or more colocaliza-

ion events. This population is referred to as colocalized objects

nd the remaining objects, containing no colocalization events, areeferred to as non-colocalized objects. To quantify the effects ofon-colocalized objects, we calculated the overall colocalizationercentage for NCX and cav-3 within just the colocalized objects,

.e. with the non-colocalized objects removed. We determined the

lso applied to the interior compartment. (A) Shows the number of interior objectsn where non-colocalized objects are excluded from the calculation. (C and D) Show) Reports the number of objects found to contain more than 50% colocalization. (F)than 50% colocalization.

median object colocalization as well as the proportion of objectsthat were highly colocalized, with more than 50% colocalization ineach object. The number of colocalization events contained in thesehighly colocalized events was also determined.

In the peripheral compartment, NCX and cav-3 appear to beminimally colocalized. Less than 40% of peripheral objects containone or more colocalization events (Fig. 5A). The percentage of NCXobjects containing colocalization events was found to be not statis-tically significant for all neighboring groups. However, percentagesof cav-3 objects containing colocalization for d3 vs. d6 and for d20

vs. d56 were found to be statistically significant.

If all NCX and cav-3 objects containing no colocalization eventsare removed which effectively restricts the region of interest to thecolocalized subpopulation, the object specific NCX and cav-3 colo-calization indices increase by about 5% (Fig. 4B vs. Fig. 5B). However,

376 E. Lin et al. / Cell Calcium 45 (2009) 369–383

Fig. 7. 3D distance analysis. Volume projection of surface connectivity between objects is shown in (A). (B) Shows a higher magnification of the connectivity grid createdb borina al NCo refera

eie

tttCntFvnvc

sciblcstir

i

y the Delaney triangulation. Each object (large green circle) can have several neighs the basis for the distance calculation. (C) Shows a volume projection of peripherutlines. For clarity (C) only shows the top half of the cell. (For interpretation of therticle.)

ven with the restricted region of interest to only include colocal-zed objects, the object specific colocalization was less than 21% forither NCX or cav-3 colocalization (Fig. 5B).

The small rise (∼5%) in overall colocalization, even withhe complete removal of all non-colocalized objects indicateshat although ∼60% of the objects were removed, indicateshat these non-colocalized objects were relatively small in size.olocalized objects were found to be roughly 3× larger thanon-colocalized objects and these differences were found to be sta-istically significant at all developmental stages (data not shown).or peripheral NCX object specific colocalization, d3 vs. d6, d6s. d10 and d20 vs. d56 were found to be statistically sig-ificant. For peripheral cav-3 object specific colocalization, d10s. d20 and d20 vs. d56 were found to be statistically signifi-ant.

Although non-colocalized objects have been removed for object-pecific colocalization calculations there remains a continuum ofolocalization indices within the remaining objects. High colocal-zation indices in individual objects reflect a high degree of overlapetween NCX and cav-3 while low colocalization indices reflect a

ow degree of overlap. Colocalized voxels that result from objectsontaining large degrees of overlap may have more physiologicalignificance than colocalized voxels resulting from objects that con-

ain only small degrees of colocalization. Therefore, colocalizationndices within individual objects were quantified as small isolatedegions of interest.

A representative histogram, from the 20 day group, is shownn Fig. 5C and D for NCX and cav-3 object colocalization, respec-

g objects (small green circles) and the connecting path between two points servesX and cav-3 objects from the prospective of the camera as shown in (A) by the redences to color in this figure legend, the reader is referred to the web version of the

tively. Histograms from the other age groups revealed a similarshape and trend in their profiles. The median object colocaliza-tion was 19% and 17% for NCX and cav-3 objects, respectively,indicating that half of the objects had colocalization indices lessthan that. The 56 day group had the largest median colocaliza-tion index, with a 22% median NCX object colocalization. The largeright skew of the cumulated histogram indicates that the major-ity of colocalization events arise from events that are minimallycolocalized.

To quantify the significance of highly colocalized objects (i.e.objects containing more than 50% colocalization) the proportionand contribution of these events was determined. Of the colocal-ized objects, each containing at least one colocalized voxel, only∼7% of peripheral colocalized objects contained 50% or greater colo-calization (Fig. 5E). Statistical tests between neighboring age groupsfound only the d3 vs. d6 comparison for the proportion of highlycolocalized NCX objects to be significantly significant. All othercomparisons found no significant differences between neighboringage groups.

The contribution of these highly colocalized objects, as in theproportion of peripheral colocalization voxels that can be attributedto these highly colocalized objects is reported in Fig. 5F. The contri-bution of these highly colocalized objects did not contribute more

than 23% of the overall colocalization events in the cell periph-ery. Except for the comparison between the contributions of highlycolocalized NCX objects between d6 vs. d10, all other comparisonsbetween neighboring age groups were found to be not statisticallysignificant.

lcium 45 (2009) 369–383 377

3

uetddippewgiotc

rritda

oNtcincFac

3

ntoit

tss1ocp3Ns

cdtdlmlfin

Fig. 8. Distance to closest neighbor. Distance analysis was used to quantify the sep-aration distances to the closest neighboring object: between two NCX objects, twocav-3 objects and from a NCX object to a cav-3 object. (A) Shows NCX to NCX separa-

E. Lin et al. / Cell Ca

.5. Object specific colocalization in the interior compartment

Identical analysis utilized for the peripheral compartment wastilized for the interior compartment. In contrast to the periph-ral population, levels of colocalization in the cell interior appearedo change with development (Fig. 6). Because significant t-tubuleevelopment only occurs after 10 days post-partum, only the d10,20 and d56 groups are considered in this section. Within the cell

nterior, NCX colocalization was greater than cav-3 colocalizationartially due to a larger amount of cav-3 voxels. Although the pro-ortion of t-tubule objects containing one or more colocalizationvents was less than in the periphery, this proportion increasedith development (Fig. 6A). From the 10 day group to the 56 day

roup, the percent NCX objects containing colocalization eventsncreased from 20% to 36%. Over the same time frame, the percentf colocalized cav-3 objects increased from 11% to 34%. Statisticalests between d10 vs. d20 and between d20 vs. d56 found thesehanges to be statistically significant.

NCX and cav-3 object specific colocalization in the cell inte-ior (i.e. interior colocalization with the non-colocalized objectsemoved) increased with age. NCX object specific colocalizationncreased from 20% to 30% while cav-3 object specific colocaliza-ion increased from 16% to 20%. Differences between the d20 vs.56 groups were found to be statistically significant for both NCXnd cav-3 object specific colocalization in the interior.

Similar to the histogram profiles of NCX and cav-3 objectsn the periphery, interior histograms indicated large right skews.ote that in the interior, there tend to be more cav-3 object

han NCX objects. The percentage of interior objects with colo-alization indices greater than 50% is shown in Fig. 6E. Thencreases in NCX and cav-3 proportions are statistically sig-ificant between d10 vs. d20 and between d20 and d56. Theontribution of the highly colocalized objects is shown inig. 6F. Comparisons between d10 vs. d20 and between d20nd d56 found that these changes were not statistically signifi-ant.

.6. Peripheral distributions of NCX and cav-3

Colocalization measurements are best suited to describingon-resolvable distributions, in which two proteins are dis-ributed within a common voxel space. The large proportion ofbjects that are either non-colocalized or only lowly colocalizedndicates many objects may be separated by a measurable dis-ance.

Fig. 8 indicates that the overall distributions of NCX and cav-3 onhe cell periphery do not appear to change between developmentaltages. Fig. 8A (closed symbols) indicates that NCX objects wereeparated from other NCX objects most often by a distance of about�m. Fig. 8B (closed symbols) indicates that the separation profilesf cav-3 objects appears to follow a similar pattern, where the mostommon separation distance is also 1 �m with no changes in therofile with development. The relationship between NCX and cav-objects is shown in Fig. 8A (open symbols) and indicates thatCX and cav-3 objects appear to be well separated with a peak

eparation distance of about 0.5 �m.To investigate these separation profiles further, we also cal-

ulated the separation profiles in the lateral and longitudinalirections. Fig. 9 indicates that lateral and longitudinal separa-ion distances for all three measurements do not appear to beevelopmentally regulated and that lateral separation profiles and

ongitudinal separation profiles are very similar between develop-ental stages. Separation profiles shown in Fig. 9 appear to be more

eft-skewed in comparison with profiles from Fig. 8. This shift tonding longer distances when determining lateral and longitudi-al separations may be a reflection of the decreased probability

tion distances for the different developmental stages in closed circles (right curve)and NCX to cav-3 separation distances in open squares (left curve). (B) Show thecav-3 to cav-3 separation distances.

of finding a lateral as well as a longitudinal neighbor for each andevery node. The lack of an intermediate neighbor allows a pseudo-neighbor to be included (the neighbor’s neighbor).

Because NCX to NCX, cav-3 to cav-3 and NCX to cav-3 separa-tion profiles did not appear to change with development, Fig. 10summarizes the overall peripheral distribution by cumulating datafrom all age groups. Fig. 10A and B summarizes the angular relation-ship between NCX and NCX objects and between cav-3 and cav-3objects, respectively. Data points along the ±90o positions corre-spond with longitudinal measurements while data points at 0o and±180o correspond with lateral measurements as per Fig. 9. About6000 events were also found at the intermediate angles at ±45o

and at ±135o. Fig. 10C provides distance profiles for these eight

directions.For the 20 day and 56 day groups, internal separation pro-

files were also determined (Fig. 11). For internal measurements,measurements were determined 3 voxels in from the cell surfacebut otherwise similar to determining peripheral measurements.Relative to measurements of peripheral separation, internal mea-surements are low probability events. Neither NCX nor cav-3 labelsappear to envelope large amounts of t-tubular membrane. As such,the distances to the closest neighboring object were often greaterthan 3 �m and are not included in the figures. Lateral separationprofiles appear similar with peripheral separation profiles with

peak separation at around 1 �m. Longitudinal separation profiles,however, are distinctly different than their peripheral counterparts.Although on the periphery, NCX-to-NCX and cav-3-to-cav-3 sepa-ration are at 1 �m, interior measurements indicate that the mostcommon separation distances are 2–3 �m in length.

378 E. Lin et al. / Cell Calcium 45 (2009) 369–383

F hborsa n distN iddleb

4

atfindatomfo

ig. 9. Peripheral lateral and longitudinal separation. Lateral and longitudinal neigxis relative to the center node. The left columns (A, C, E) indicate lateral separatioCX to NCX (A, B) and cav-3 to cav-3 (C, D) distances are shown in the top row and metween NCX and cav-3 objects.

. Discussion

Thresholding criteria are an integral part of every imaging studynd in this study we utilized an iterative thresholding techniquehat determined a threshold value based on the number of objectsound at each threshold value between 0 and 10,000. To avoid elim-nating dim objects, the threshold that generated the maximumumber of events was used. Over thresholding was checked for byetermining the median object size as a function of polar anglend it was determined that objects along the tops and bottoms of

he cells were smaller than objects on the left and right aspectsf the cell. Although objects were smaller on the tops and bottoms,edian object size was still greater than 10 voxels per object. There-

ore, this iterative technique did not appear to result in systematicver thresholding.

were defined as objects that were located ±15% from the lateral and longitudinalances while the right columns (B, D, F) indicate longitudinal separations distances.

row, respectively. (E and F) Show the lateral and longitudinal separation distances

Several stages of colocalization analysis were employed in thisstudy, ranging from whole cell measurements to individual objectmeasurements and we did not find the majority of NCX and cav-3to be colocalized at any age group. Colocalization at the whole celllevels was less than 15% colocalization for all age groups. Analy-sis of the peripheral and interior compartments indicated that themajority of colocalization events occurred on the periphery. Devel-opment brings about increased colocalization in both the peripheraland interior compartments with larger changes occurring in theinterior. Statistical analysis reveals that changes in colocalization

between 3 day vs. 6 day vs. 10 day vs. 20 day groups are notstatistic significant. However, the 56 day group was found to bestatistically different from all other groups. Subsequent removal ofnon-colocalized objects (object specific colocalization) increasedpeak levels of peripheral colocalization to 21% and peak levels of

E. Lin et al. / Cell Calcium 45 (2009) 369–383 379

Fig. 10. Directionality of neighboring objects. Distances to the closest neighbor (Fig. 8) as well as lateral and longitudinal separation distance profiles (Fig. 9) suggested thatthe distribution patterns of NCX and cav-3 objects on the periphery do not change with development. To take advantage of this property, object neighbor data were pooled andthe angular relationship between neighbors was derived. (A and B) Show the relative angular relationship between NCX neighbors and cav-3 neighbors, respectively. Anglel dinal dt me ass retativ

idw

tiS3ranmw

abels represent the direction of the neighbor relative the center object. The longituhese graphs. (C) Shows cumulated data from all 5 age groups and is oriented the sahown in green. NCX to cav-3 separation distances are shown in yellow. (For interpersion of the article.)

nterior colocalization to 30%. Interior colocalization increased withevelopment and myocytes may continue to mature after 56 dayshich may result in further increases in colocalization.

The role of cav-3 in E–C coupling seems unclear given data fromhis and several other studies. Both imaging and biochemical stud-es have reported differing relationships between NCX and cav-3.criven et al. showed that on the surface of adult rat myocytes,0–40% of voxels labeled for CaV1.2, the sodium channel isoform

H1, NCX and RyR also contained cav-3. Since Scriven et al. [21] hadlready reported that NCX and RyR, and NCX and Na+ channel doot colocalize in adult rat cardiomyocytes, this suggested that thereust be several subpopulations of cav-3 with varying associationsith other proteins.

irection is oriented vertically while the lateral direction is oriented horizontally in(A) and (B). NCX to NCX distances are shown in red and cav-3 to cav-3 distances areon of the references to color in this figure legend, the reader is referred to the web

The implications of their NCX and cav-3 colocalization datasuggest that the surface distributions of NCX and cav-3 in therat cardiomyocyte may be fundamentally different from that inthe rabbit. In our study, most cells had a similar number of NCXand cav-3 voxels on the cell periphery, an effect that is visible inthe similarities between colocalization values for that compart-ment (Fig. 4). Because NCX and cav-3 colocalization calculationsshare the same numerator, similar colocalization percentages indi-

cates a similar sized denominator. Scriven et al. [19] found 30%NCX colocalization with cav-3 and 10% cav-3 colocalization withNCX which indicates that there is 3 times more cav-3 voxelsthan NCX voxels on the surface of adult rat myocytes. Whetherthis difference implicates differing distribution periodicities or

380 E. Lin et al. / Cell Calcium

Fig. 11. Interior separation distances. Separation distances were also determinedfgsc

dk

prufmiacfpbsaro

qortpEcmotcc

rom NCX and cav-3 objects directly beneath the peripheral compartment. The lon-itudinal separation for NCX to NCX is shown with square symbols while the lateraleparation is shown with triangle symbols. The longitudinal separation of cav-3 toav-3 is shown with diamonds while the lateral separation is shown with circles.

iffering objects sizes between rat and the rabbit is not yetnown.

Bossuyt et al. [18] found that NCX and cav-3 co-immuno-recipitate in bovine cardiac sarcolemmal vesicles. However, aecent study by Cavalli et al. [20] reported NCX and cav-3 to ben-associated by both imaging and biochemical techniques. Theyound NCX and caveolin to stain the surface and interior compart-

ents strongly but colocalization correlation coefficients did notndicate significant colocalization. The relationship between NCXnd cav-3 was also tested using differing lipid raft isolation proto-ols. Using a “Na2CO3, pH 11” protocol to isolate rafts, this protocolound an association between NCX and cav-3. However, when therotocol was changed to a “Triton X-100” protocol, an associationetween NCX and cav-3 was no longer observed. While biochemicaltudies are able to make inferences about actual protein associ-tions under different environments or isolation protocols, theseesults may not reflect differing cav-3 subpopulations that mayccur physiologically.

We determined the overall NCX and cav-3 colocalization to beuite low but also found a small population of NCX and cav-3bjects that are highly colocalized. About 6% of peripheral and inte-ior objects were found to have a colocalization index of greaterhan 50%. However, the functional contributions of these smallopulations are unknown. cav-3 is likely to participate in both–C coupling and non-E–C coupling related roles therefore; overallellular colocalization between the protein in question and cav-3

ay be difficult to interpret in conditions in which the number

f cav-3 voxels is large. In this study, we partially controlled forhe effects of lowly colocalized cav-3 voxels by performing colo-alization on an object-by-object basis, thereby separating highlyolocalized events from the non-colocalized population. In protein

45 (2009) 369–383

distributions that contain distinct subpopulations, the colocal-ization indices of individual objects may be more reflective ofphysiological coupling than a global colocalization measure.

4.1. Protein separation distances

The distance data suggest that NCX and cav-3 are not rigidly dis-tributed on the periphery of the cell since separation distances varyconsiderably. However, NCX and cav-3 are likely to be distributedwith an underlying pattern that does not appear to change duringdevelopment. Since the majority of NCX and cav-3 objects do notappear to be colocalized, most of NCX and cav-3 are distributed in amutually exclusive pattern. Closest neighbor analysis indicates thatNCX is often within 0.5 �m of a cav-3 cluster (and vice versa) andthat both NCX and cav-3 are separated from other NCX and cav-3objects often by ∼1.0 �m. Angle analysis (Fig. 10) indicates that NCXand cav-3 objects are likely to have eight neighboring objects of thesame type at every half turn or every 45o. Four neighbors are locatedlongitudinally and laterally (positions ±90◦, ±180◦) and four neigh-bors located diagonally (positions ±45◦, ±135◦). This may suggestthat NCX and cav-3 objects are organized in a grid-like manner.Internal separation distances revealed that NCX and cav-3 have asimilar lateral separation profile to peripheral measurements buthave a 2-fold longer longitudinal periodicity in the cell interior ver-sus the cell surface. The vast majority of interior objects were notcontiguous with surface objects and we were unable to registersurface events with t-tubule and z-line positions.

4.2. Summary

In this study NCX and cav-3 objects categorized into non-colocalized (0%), lowly colocalized (<50%) and highly colocalized(> 50%) subpopulations in both the interior and peripheral com-partments. The majority of colocalization events were found to beassociated with lowly colocalized objects but a small and possibilityphysiologically important population of highly colocalized eventswas also found. Distance analysis revealed that the distribution ofNCX and cav-3 does not appear to vary with development and wedemonstrate for the first time that these two proteins appear to bedistributed in an alternating checker pattern with a common lateraland longitudinal periodicity.

Acknowledgements

The insightful input of Dr. Ed Moore at the University of BC isgratefully acknowledged. This work was supported by the Canadian

Institutes of Health Research (to GFT). V. Hung was the recipientof a Hardwick Graduate Studentship from the University of BritishColumbia. G.F. Tibbits is the recipient of a Tier I Canada ResearchChair. None of the authors has a commercial interest or conflict ofinterest in this study.

E.Linet

al./CellCalcium45

(2009)369–383

381

Appendix A

382E.Lin

etal./CellCalcium

45(2009)

369–383

Appendix B

lcium 45 (2009) 369–383 383

A

R

[

[

[

[

[

[

[

[

[

[

[20] A. Cavalli, M. Eghbali, T.Y. Minosyan, E. Stefani, K.D. Philipson, Localization of

E. Lin et al. / Cell Ca

ppendix C

eferences

[1] D.M. Bers, Calcium cycling and signaling in cardiac myocytes, Annu. Rev. Physiol.70 (2008) 23–49.

[2] J. Huang, L. Hove-Madsen, G.F. Tibbits, Ontogeny of Ca2+-induced Ca2+ release inrabbit ventricular myocytes, Am. J. Physiol. Cell Physiol. 294 (2008) C516–C525.

[3] M. Artman, H. Ichikawa, M. Avkiran, W.A. Coetzee, Na+/Ca2+ exchange currentdensity in cardiac myocytes from rabbits and guinea pigs during postnataldevelopment, Am. J. Physiol. 268 (1995) H1714–H1722.

[4] S.R. Boerth, D.B. Zimmer, M. Artman, Steady-state mRNA levels of the sarcolem-mal Na+–Ca2+ exchanger peak near birth in developing rabbit and rat hearts,Circ. Res. 74 (1994) 354–359.

[5] M. Artman, Sarcolemmal Na+–Ca2+ exchange activity and exchanger immunore-activity in developing rabbit hearts, Am. J. Physiol. 263 (1992) H1506–H1513.

[6] R. Vetter, R. Studer, H. Reinecke, F. Kolar, I. Ostadalova, H. Drexler, Reciprocalchanges in the postnatal expression of the sarcolemmal Na+–Ca2+-exchangerand SERCA2 in rat heart, J. Mol. Cell Cardiol. 27 (1995) 1689–1701.

[7] M.D. Harrell, S. Harbi, J.F. Hoffman, J. Zavadil, W.A. Coetzee, Large-scale analysisof ion channel gene expression in the mouse heart during perinatal develop-ment, Physiol. Genom. 28 (2007) 273–283.

[8] J. Huang, L. Hove-Madsen, G.F. Tibbits, Na+/Ca2+ exchange activity in neonatalrabbit ventricular myocytes, Am. J. Physiol. Cell Physiol. 288 (2005) C195–203.

[9] G.T. Wetzel, F. Chen, T.S. Klitzner, L- and T-type calcium channels in acutelyisolated neonatal and adult cardiac myocytes, Pediatr. Res. 30 (1991) 89–94.

10] J. Huang, L. Xu, M. Thomas, K. Whitaker, L. Hove-Madsen, G.F. Tibbits, L-typeCa2+ channel function and expression in neonatal rabbit ventricular myocytes,Am. J. Physiol. Heart Circ. Physiol. 290 (2006) H2267–H2276.

11] F. Sedarat, E. Lin, E.D. Moore, G.F. Tibbits, Deconvolution of confocal imagesof dihydropyridine and ryanodine receptors in developing cardiomyocytes, J.Appl. Physiol. 97 (2004) 1098–1103.

[

12] G.T. Lines, J.B. Sande, W.E. Louch, H.K. Mork, P. Grottum, O.M. Sejersted, Con-tribution of the Na+/Ca2+ exchanger to rapid Ca2+ release in cardiomyocytes,Biophys. J. 91 (2006) 779–792.

13] J.S. Frank, G. Mottino, D. Reid, R.S. Molday, K.D. Philipson, Distribution of theNa+–Ca2+ exchange protein in mammalian cardiac myocytes: an immunoflu-orescence and immunocolloidal gold-labeling study, J. Cell Biol. 117 (1992)337–345.

14] F. Chen, G. Mottino, T.S. Klitzner, K.D. Philipson, J.S. Frank, Distribution of theNa+/Ca2+ exchange protein in developing rabbit myocytes, Am. J. Physiol. 268(1995) C1126–C1132.

15] P. Dan, E. Lin, J. Huang, P. Biln, G. Tibbits, Three dimensional distribution of car-diac Na+–Ca2+ exchanger and ryanodine receptor during development, Biophys.J. 93 (2007) 2504–2518.

16] B. Razani, S.E. Woodman, M.P. Lisanti, Caveolae: from cell biology to animalphysiology, Pharmacol. Rev. 54 (2002) 431–467.

17] T.M. Williams, M.P. Lisanti, The caveolin proteins, Genome Biol. 5 (2004)214.

18] J. Bossuyt, B.E. Taylor, M. James-Kracke, C.C. Hale, Evidence for cardiacsodium-calcium exchanger association with caveolin-3, FEBS Lett. 511 (2002)113–117.

19] D.R. Scriven, A. Klimek, P. Asghari, K. Bellve, E.D. Moore, Caveolin-3 is adja-cent to a group of extradyadic ryanodine receptors, Biophys. J. 89 (2005)1893–1901.

sarcolemmal proteins to lipid rafts in the myocardium, Cell Calcium 42 (2007)313–322.

21] D.R. Scriven, P. Dan, E.D. Moore, Distribution of proteins implicated inexcitation–contraction coupling in rat ventricular myocytes, Biophys. J. 79(2000) 2682–2691.