Research ArticleVASCULAR BIOLOGY

Spatially structured cell populations process multiple sensory signals in parallel in intact vascular endothelium

See allHide authors and affiliations

Science Signaling  18 Dec 2018:
Vol. 11, Issue 561, eaar4411
DOI: 10.1126/scisignal.aar4411

Heterogeneity and cooperation in the endothelium

Endothelial cells lining blood vessels must sense, interpret, and respond to many different chemical signals. Coordination of the responses of these cells is critical for the proper function of the cardiovascular system. Lee et al. found that the coordinated behavior of endothelial cells in rat carotid artery resulted from communication between heterogeneous populations of cells with differential sensitivities to biochemical cues. Spatially distinct clusters of cells responded to either muscarinic or purinergic agonists; few cells were responsive to both. Each agonist evoked distinct intracellular signals, but communication between the different cell clusters generated new composite signals when both agonists were present. These results contribute to understanding how the endothelium can process large amounts of biochemical information for a coordinated, tissue-wide response.

Abstract

Blood flow, blood clotting, angiogenesis, vascular permeability, and vascular remodeling are each controlled by a large number of variable, noisy, and interacting chemical inputs to the vascular endothelium. The endothelium processes the entirety of the chemical composition to which the cardiovascular system is exposed, carrying out sophisticated computations that determine physiological output. Processing this enormous quantity of information is a major challenge facing the endothelium. We analyzed the responses of hundreds of endothelial cells to carbachol (CCh) and adenosine triphosphate (ATP) and found that the endothelium segregates the responses to these two distinct components of the chemical environment into separate streams of complementary information that are processed in parallel. Sensitivities to CCh and ATP mapped to different clusters of cells, and each agonist generated distinct signal patterns. The distinct signals were features of agonist activation rather than properties of the cells themselves. When there was more than one stimulus present, the cells communicated and combined inputs to generate new distinct signals that were nonlinear combinations of the inputs. Our results demonstrate that the endothelium is a structured, collaborative sensory network that simplifies the complex environment using separate cell clusters that are sensitive to distinct aspects of the overall biochemical environment and interactively compute signals from diverse but interrelated chemical inputs. These features enable the endothelium to selectively process separate signals and perform multiple computations in an environment that is noisy and variable.

INTRODUCTION

The vascular endothelium is a continuous network of about 10 trillion cells (1) that controls virtually all cardiovascular behavior. It is the sensory interface for an enormous quantity of information about the chemical environment to which the vascular system is exposed and which provides cues on physiological status. Blood composition, hormones, neurotransmitters, endothelial cells, pericytes, smooth muscle cells, various blood cells, viral or bacterial infection, and proinflammatory cytokines provide numerous signals that instruct the vascular system. Many (minimally tens) of these extracellular signals may arrive simultaneously, often fluctuating around basal concentration values creating multiple, small extracellular signals that are difficult to resolve (25). The endothelium processes the information conveyed by each chemical signal and in the range of concentrations over which each signal fluctuates. The endothelium must manage its responses to multiple extracellular signals simultaneously, requiring selective detection, processing, and ultimately, integration of separate inputs.

Although heterogeneity in the endothelium is known to exist (69), the endothelium is usually treated as a homogeneous population of cells that responds uniformly to each activator. It has often been assumed that biological systems maximize tissues’ ability to respond to perturbations through the coordinated responses of homogeneous populations of cells (10). The classical view of endothelial function is that information arriving to each cell is interpreted and conveyed to neighboring cells (such as other endothelial cells or smooth muscle cells), much like a cable, without changing the information content (11, 12). Given this proposed arrangement, precisely how the endothelium conveys multiple independent extracellular signals arriving simultaneously to elicit particular cell and tissue responses is not clear.

Chemical stimuli that activate endothelial cells are often transduced as changes in cytosolic Ca2+ concentration (1319), which may act alone or cooperate with other factors to elicit cellular responses (20). Ca2+ links extracellular stimuli to physiological responses by regulating the synthesis and release of various vasoactive agents such as nitric oxide (NO), various peptides, and eicosanoids such as prostacyclin and thromboxane [reviewed in (21)]. Through these Ca2+-dependent mediators, the endothelium controls vascular tone, nutrient exchange, blood cell recruitment, blood clotting, and the formation of new blood vessels [reviewed in (21)]. Cardiovascular function requires careful targeting of the physiological information provided by extracellular activators, through changes in intracellular Ca2+, to single-cell activities and the physiological behavior of the entire endothelium. Targeting is successful when the activity of a specific intracellular process alters in response to a Ca2+ signal. Specific targeting of particular intracellular processes by extracellular signals is believed to rely mainly on the amplitude or temporal or spatial features of the Ca2+ signal within each cell (2225). However, despite the diversity of signals, various activators evoke Ca2+ increases that often appear similar in their localized increases, uniform global rises, or propagating waves that occur through all or part of a cell (2629). How these Ca2+ signals selectively evoke a multitude of distinct physiological responses is not yet fully understood.

When several chemical mediators are present simultaneously, the Ca2+ signal must integrate information from multiple sources while still producing distinct signals to control multiple activities, thus presenting an additional complex sensory problem. An appreciation of how Ca2+ selectively couples different sensory inputs in a complex chemical environment to various physiological functions is therefore central to understanding endothelial function.

To control cardiovascular function, the endothelium extracts information from a noisy and variable environment by performing multiple simultaneous detections of chemical signals through changes in Ca2+ concentration to monitor physiological status. A key to an understanding of this crucial aspect of the cardiovascular function are the mechanisms by which endothelial cells integrate information from the local chemical environment. To address this issue, we recorded the concurrent activity from hundreds of endothelial cells in intact blood vessels and examined the responses of individual cells to extracellular activators of muscarinic and purinergic receptors. The response to muscarinic receptor activation is thought to underlie the response to hypothermia (30) and shear stress (31). Purinergic stimulation may occur by adenosine triphosphate (ATP) released from activated platelets (32, 33) to evoke vasodilation (34) and either an increase (35) or decrease (36) in endothelial permeability.

Our results show that the endothelium is organized into spatially structured clusters of endothelial cells that respond to a given stimulus. The clusters are sensitive to particular stimuli and separately process various elements of the chemical environment in parallel. Cells within clusters share information with neighbors. When more than one agonist is present, cells perform computations to generate new signals that are distinct from those elicited by the individual signals. The endothelium thus appears to act as a network of structured, collaborating sensors that act in parallel to interpret the multiple signals that report physiological status.

RESULTS

Heterogeneous Ca2+responses of endothelial cells to carbachol and shear stress

The endothelium is a distributed sensory system that can detect a large number of chemical inputs (21, 37). However, it is not known whether the distribution of sensory cells is uniform, random, or clustered. As a first step in determining the distribution of sensory cells and how the endothelium responds to extracellular activators, we evaluated agonist-induced Ca2+ signaling in large numbers (~200) of endothelial cells in intact arteries.

The muscarinic agonist carbachol (CCh) evoked an increase in Ca2+ concentration in several cells in the endothelium (Fig. 1A and fig. S1, A to D). The cells that activated Ca2+ release appeared to form clusters within the endothelium (movie S1). Each active cell’s response consisted of Ca2+ waves that propagated across all or part of the cell (movie S1). Some cells were inactive or barely active, and others were highly active, with repeating Ca2+ waves and oscillations (Fig. 1, A to F). The amplitude and duration of the responses also varied across cells (Fig. 1, C and D). The differences in amplitudes, duration, and oscillation frequency from individual cells gave rise to a wide range of distinct responses (Fig. 1, A, C, and D). There also was considerable heterogeneity in the Ca2+ response to a different stimulus, fluid flow (shear stress), in the endothelium (fig. S2, A to F). Despite the wide variation in amplitudes, durations, and oscillations in individual cells, when the response to CCh was averaged across all cells or the entire field of view, a monophasic rise in Ca2+ occurred (fig. S1, E and F).

Fig. 1 CCh-evoked Ca

2+ signals in the endothelium of intact arteries. (A) A representative image frame showing ~200 endothelial cells from an imaging field with automatically identified regions of interest (ROIs; red outlines) superimposed onto the imaging field of view. Ca2+ imaging (green) illustrates the endothelial response to treatment with CCh (100 nM) overlaid on a basal (no treatment) Ca2+ image. (B) Expanded view of the endothelium from the area highlighted in the red box in (A). Seven different ROIs are noted with different color outlines. (C) Baseline-corrected Ca2+ signals (F/F0) obtained from the fluorescence within the ROIs shown in (B). The color for each trace matches that for the ROI in (B). (D) Overlaid Ca2+ signals from all individual cells identified in (A). The black line shows the mean response. (E) A representative single endothelial cell Ca2+ signal and (F) the corresponding images with time points indicated by the roman numerals. Scale bars, 50 μm. Data are representative of n = 50 independent experiments from different artery preparations.

CCh-evoked changes in endothelial Ca2+ were required for relaxation of the arteries. When Ca2+ changes in the endothelium were buffered by perfusing the chelator 1,2-bis(2-aminophenoxy)ethane-N,N,N′,N′-tetraacetic acid (BAPTA) into the artery lumen before CCh stimulation, the relaxation evoked by a maximal concentration of CCh was inhibited (fig. S3, A to C).

Concentration dependence of cholinergic signaling

Previous studies have shown the averaged or ensemble endothelial Ca2+ response to various stimuli increases in a concentration-dependent manner (16, 38). However, it is not clear whether these increases in response arose from an increase in the Ca2+ response in each single cell or an increase in the number of cells being activated, with each responding with a maximum response, because single-cell concentration-response relationships are rarely carried out. We found that the number of cells responding with Ca2+ releases increased with CCh concentration (Fig. 2A). The cells that activated at each concentration appeared to be positioned in clusters in various areas of the endothelium. The range [10% effective concentration (EC10) to EC90] of CCh concentrations over which cells responded was 123 nM to 1.2 μM (95% confidence intervals, 76 to 199 nM for EC10 and 740 nM to 2.1 μM for EC90; Fig. 2, B to E). These data indicate that at least part of the concentration dependence of the response to muscarinic activation arises from increased recruitment of responding endothelial cells.

Fig. 2 Concentration dependence of CCh-evoked endothelial Ca

2+ signaling. (A) Representative composite images showing endothelial Ca2+ activity in rat carotid artery endothelium in response to increasing concentrations of CCh. The three images show illustrative data from a single experiment in which the endothelium was stimulated with increasing concentrations of CCh consecutively. The full CCh concentration series applied in each experiment was 10 increments from 1 nM to 30 μM. Scale bars, 50 μm. (B) Average Ca2+ responses (F/F0) from all cells for each concentration of the complete CCh concentration series. The individual traces are color-coded on the basis of the concentration of CCh applied. (C) Ca2+ traces from a single cell. The color code is the same as that in (B). (D) Percentage of cells activated by increasing concentrations of CCh. The x axis is log10 of the CCh concentration. M, molar. (E) Ca2+ traces from all 129 individual cells from the dataset shown in (A) to (D). The black line represents the mean response. Different colors were assigned to the traces based on the amplitude of the initial response to CCh in each cell at the highest CCh concentration (far right). The color of each specific cell’s trace was maintained throughout subsequent additions of agonist, thus allowing for direct comparison of the same cells between different CCh concentrations. Data are representative of n = 5 independent experiments from different artery preparations from different animals.

In addition to the increased recruitment of responding cells, the amplitude of the Ca2+ response in each cell also increased in a concentration-dependent manner (Fig. 2, C and E). The EC50 of CCh for endothelial cell activity was 630 nM (95% confidence interval, 399 nM to 996 μM; fig. S4A). The range (EC10 to EC90) over which the magnitude of the Ca2+ response of responding cells increased was 50 nM to 8 μM (95% confidence intervals, 17 to 142 nM for EC10 and 2.9 to 22.1 μM for EC90; fig. S4B). Thus, the concentration-dependent response was a combination of increasing recruitment of cells and increasing amplitude of responses within cells (fig. S4, A to D).

There were differences in the sensitivities of concentration responses in amplitude and number of cells activated (Fig. 2D and fig. S4B). All cells could be activated before a maximal response was achieved, as expected from a population of cells with various sensitivities. A composite measure (fig. S4B) of the percentage of cells activated multiplied by the average amplitude of the initial Ca2+ rise in responding cells at each concentration showed that the EC50 of CCh required for total endothelial activation (all cells in the endothelium responding) was 1.1 μM (95% confidence interval, 743 nM to 1.5 μM). The range (EC10 to EC90) of CCh concentrations over which cells responded was 164 nM to 6.8 μM (95% confidence intervals, 76 to 354 nM for EC10 and 3.2 to 14.8 μM for EC90). The concentration-response relationship for total endothelial response was significantly right-shifted compared to the relationship describing percentage of cell activation (fig. S4B versus Fig. 4D). Thus, the concentration dependence of the response of individual cells increased the overall sensitivity of the endothelium.

The Ca2+ increase evoked by CCh was generated through muscarinic receptors, as confirmed by the complete block of the response by the M3 receptor antagonist 4-DAMP (1,1-dimethyl-4-diphenylacetoxypiperidinium iodide) (Fig. 3, A to E). Whereas repeated stimulation with CCh at 10-min intervals evoked reproducible responses (Fig. 3, A to D), the addition of the M3 receptor antagonist (4-DAMP, 100 nM) blocked subsequent responses to CCh (Fig. 3, A to E). This block was not due to reduced viability of the cells because the cells were still able to respond to the purinergic receptor agonist ATP (Fig. 3, A to D).

Fig. 3 CCh activates M3 receptors in the endothelium.

(A) Representative experiment showing the effect of the selective M3 antagonist 4-DAMP on the average maximal CCh-evoked Ca2+ activity across all cells in an endothelial sample. Two responses to CCh were measured before the addition of 4-DAMP. CCh and 4-DAMP applications are indicated by the lines above the traces. After CCh and 4-DAMP were washed out, cells were stimulated with ATP to test for viability. (B) Composite Ca2+ images showing the cells responding to two consecutive applications of CCh (green), followed by an application of CCh plus 4-DAMP. Cell viability was assayed by measuring the Ca2+ response to ATP (red). All images are from the same field of endothelium. Scale bars, 50 μm. (C) Ca2+ signals from each cell in (B), including 30 s of baseline readings and 60 s of activation. Each individual cell’s response is color-coded and overlaid. The black line represents the average. (D) Paired peak Ca2+ (F/F0) responses from individual cells (circles) matched for each treatment (gray lines) from a single experiment. The average response is indicated by white circles and matched across each treatment by a black line. The right y axis (plotting density) shows the color code used on each of the plotted data points. The axis color codes the frequency of occurrence of specific peak F/F0 values. Red indicates a higher frequency of occurrence of a particular peak F/F0 value, and blue indicates a low frequency of occurrence of a peak F/F0 value. (E) Paired summary data illustrating changes in peak F/F0 response values averaged across all cells (n = 5). *P < 0.05. Data are representative of n = 5 independent experiments from different artery preparations from different animals.

Concentration dependence of purinergic signaling

In the next series of experiments, we examined the response of individual endothelial cells to the purinergic receptor agonist ATP (Fig. 4, A to E, and fig. S5, A to D). Ca2+ responses, averaged across all endothelial cells within the field of view, showed a concentration-dependent increase in the magnitude of response (Fig. 4B). The increasing response to ATP was, similar to CCh, derived from an increased number of cells that were activated and an increased amplitude of response in each cell (Fig. 4, A to E). The cells that activated to each concentration of ATP appeared to be positioned in clusters in various areas of the endothelium (Fig. 4A).

Fig. 4 Concentration dependence of ATP-evoked endothelial Ca

2+ signaling. (A) Representative composite images showing Ca2+ activity in rat carotid artery endothelium in response to stimulation with increasing concentrations of ATP. The three images show illustrative data from a single experiment in which the endothelium was stimulated with increasing concentrations of ATP (from 100 nM to 1 mM) consecutively. Scale bars, 50 μm. (B) Average Ca2+ responses (F/F0) responses from all cells for each concentration of the complete ATP concentration series. The individual traces are color-coded on the basis of the concentration of ATP applied. (C) Ca2+ traces from a single cell. The color code is the same as that in (B). (D) Percentage of cells activated by increasing concentrations of ATP. The x axis is the log10 of the ATP concentration. (E) Ca2+ traces from all 153 individual cells from the data shown in (A) to (D). The black line represents the mean response. Different colors were assigned to each trace based on the amplitude of the initial response to ATP in each cell at the highest ATP concentration (far right) to allow for direct comparison of the same cells between different ATP concentrations. Data are representative of n = 5 independent experiments from different artery preparations from different animals.

The EC50 of ATP for cell activation was 3.17 μM (95% confidence interval, 2.19 to 4.58 μM; Fig. 4D). The range (EC10 to EC90) of concentration of ATP over which cells responded was 438 nM to 22.9 μM (95% confidence intervals, 196 to 979 nM for EC10 and 10.2 to 51.2 μM for EC90; fig. S5C). When the magnitude of the peak response was considered, the EC50 of ATP was 20.5 μM (95% confidence interval, 12.2 to 34.3 μM; fig. S5A). The range (EC10 to EC90) for the magnitude of the peak Ca2+ response was 1.9 to 224 μM (95% confidence intervals, 590 nM to 5.9 μM for EC10 and 71.2 to 700 μM for EC90; fig. S5A).

When a composite measure of the percentage of cells activated multiplied by the average amplitude of the initial Ca2+ rise was assessed (fig. S5B), the EC50 of ATP for total endothelial activation was 21.4 μM (95% confidence interval, 13.9 to 33.0 μM). The range (EC10 to EC90) of concentration of ATP over which cells responded was 2.6 to 174 μM (95% confidence intervals, 1.0 to 6.7 μM for EC10 and 68.1 to 488 nM for EC90; fig. S5B). The concentration-response relationship for the total endothelial response was, similar to CCh, significantly right-shifted compared to the relationship describing percentage of cell activation (compared fig. S5B with Fig. 4D). Thus, as with muscarinic activation, although there is only a narrow range of concentrations over which the endothelium as a whole responds to ATP, the concentration dependence of the response of individual cells to purinergic activation increased the overall sensitivity of the endothelium.

Activation of purinergic receptors by ATP evoked relaxation of intact arteries (fig. S3, A to C). The relaxation evoked by ATP, similar to that evoked by CCh, required Ca2+ changes in the endothelium. When Ca2+ changes were buffered in the endothelium by perfusing the chelator BAPTA into the artery lumen before ATP stimulation, the relaxation evoked by a maximal concentration of ATP was blocked (fig. S3, A to C).

The endothelial response to ATP was mainly mediated by purinergic P2Y2 receptors rather than by P2Y1 or P2X receptors. P2Y receptors are ATP-binding heterotrimeric guanine nucleotide–binding protein (G protein)–coupled receptors that stimulate intracellular signaling events that result in the release of Ca2+ from intracellular stores. P2X receptors, on the other hand, are ATP-gated cation channels that mediate the influx of extracellular Ca2+ in response to ligand binding. Two observations support the involvement of P2Y2 receptors in the endothelial response to ATP in our system. First, the initial Ca2+ rise in response to a maximal concentration of ATP (EC25) was unaltered when extracellular Ca2+ was removed (Fig. 5, A to E), indicating that P2X receptors were not required. However, the later sustained response to ATP was reduced in a Ca2+-free bath (Fig. 5, F to J), indicating that Ca2+ influx was required for this part of the ATP-evoked Ca2+ rise. The second line of support for P2Y2 receptor involvement is that both the initial and sustained responses to a maximal concentration of ATP were largely blocked by the selective P2Y2 receptor antagonist ARC118925 (0.61 ± 0.13 F/F0 versus 0.13 ± 0.06 F/F0; Fig. 6, A to E). The P2Y1 receptor antagonist MRS2179 blocked the small residual component of the response to ATP after the addition of ARC118925 (Fig. 6, A to E). MRS2179 alone had only a small inhibitory effect on the response to ATP (66% of controls; 0.61 ± 0.13 F/F0 versus 0.4019 ± 0.09). The nearly complete block of the response to ATP by the P2Y2 antagonist suggests that the sustained part of the response involved store-operated Ca2+ entry to replenish intracellular stores rather than P2X receptor activation.

Fig. 5 Responses to ATP persist in the absence of extracellular Ca

2+. (A) Composite Ca2+ images showing Ca2+ responses (red) to two consecutive applications of ATP (EC25) in Ca2+-containing buffer (control and repeat 1), followed by application of ATP in Ca2+-free buffer [physiological saline solution (PSS)]. Ca2+ was then restored to the bathing solution for a final application of ATP (repeat 2). All images are from the same field of endothelium. (B) Ca2+ signals from the individual responding cells in (A). Twenty seconds of baseline measurements and 60 s of recording after ATP activation are shown. All traces from individual cells are overlaid, and the black line represents the average. Time of activation is shown above the trace. (C) Peak Ca2+ response (F/F0) from individual cells (circles) matched for each treatment (gray lines) from a single experiment. The average response is indicated by white circles and matched across each treatment by a bold black line. The plotting density color coding indicates the distribution of peak F/F0 values. (D) Late (steady state) Ca2+ response (F/F0) to ATP (EC25) measured at 45 s in controls, after removal and replenishment of extracellular Ca2+. (E) Average peak and steady-state responses to ATP (EC25) in control (red) and after removal of external Ca2+ (blue); n = 5, P > 0.05. (F) Composite Ca2+ images showing the Ca2+ responses to a maximal concentration of ATP (100 μM). Stimulations, removal of extracellular Ca2+, and restoration of extracellular Ca2+ were performed as in (A). All images are from the same field of endothelium. (G) Ca2+ signals from the individual responding cells in (F). All traces from individual cells are overlaid, and the black line represents the average. (H) Peak Ca2+ response (F/F0) from each individual cell (circles) matched for each treatment (gray lines) from a single experiment. Average response indicated by white circles and matched across each treatment by bold black line. The plotting density color coding indicates the distribution of peak F/F0 values. (I) Late (steady state) Ca2+ response (F/F0) to ATP (maximal concentration; 100 μM) measured at 45 s in controls, after removal of external Ca2+, and after the return of extracellular Ca2+. (J) Average peak and steady-state responses to ATP (maximal concentration; 100 μM) in control (red) and after removal of external Ca2+ (blue); P > 0.05 for the peak response; *P < 0.05 for the late response. All scale bars, 50 μm. Data are representative of n = 5 independent experiments from different artery preparations from different animals.

Fig. 6 ATP activates P2Y2 receptors in the endothelium.

(A) Representative experiment showing the effect of the P2Y1 antagonist MRS2179 and P2Y2 antagonist ARC118925 on the maximal Ca2+ response to ATP. The experiments were conducted consecutively on the same sample, and the traces show the average responses from all cells in the field. At the end of the experiment, CCh was applied to confirm cell viability. (B) Composite images showing the Ca2+ responses of cells to ATP in the presence of MRS2179, ARC118925, or both (red) and the subsequent responses to CCh (green) at the end of the experiment. All images are from the same field of endothelium. Scale bars, 50 μm. (C) Ca2+ signals from each cell in (B), including 30 s of baseline and 60 s of activation. (D) Peak Ca2+ response (F/F0) from each individual cell (circles) matched for each treatment (gray lines) from a single experiment. The average response is indicated by white circles and matched across each treatment by a bold black line. The plotting density y axis shows the color code used on each of the plotted data points. The axis color codes the frequency of occurrence of specific peak F/F0 values. Red indicates a higher frequency of occurrence of a particular peak F/F0 value, and blue indicates a low frequency of occurrence of a peak F/F0 value. (E) Paired summary data illustrating changes in peak F/F0 response values averaged across all cells. Each color represents a different animal and is maintained across the different treatments (n = 5). *P < 0.05. Data are representative of n = 5 independent experiments from artery preparations from different animals.

Distribution of sensitivities among cells to puringeric and muscarinic activation

We next addressed the distribution of cells that were most sensitive to CCh and ATP. The concentration at which 25% and 75% of the cells responded to each agonist (EC25 and EC75) was obtained from the concentration-response curves (Figs. 2D and 4D) and used to determine which cells were activated by low and high concentrations of each agonist. The experiments examining each agonist were carried out on precisely the same preparation and field of endothelial cells and show that each agonist activated a distinct subset of cells. In examining the different populations of cells activated by each agonist, we carried out two analyses. In the first analysis, we restricted measurements to the first 4 s of activation (“first-responding cells”) to minimize the effects of Ca2+ waves moving from an activated cell to neighboring cells through gap junctions (39). In the second analysis, we measured the response of all activated cells at 60 s (“full response”).

We examined which cells were activated by the EC25 and EC75 of CCh and ATP in a single field of endothelium (Fig. 7A). A low concentration of CCh (EC25) produced an initial response (first-responding cells) in a small percentage (2% to 7%) of endothelial cells (Fig. 7A). Additional cells were subsequently recruited (full response) either by direct activation by the agonist or by the Ca2+ increase in the initial responding cells propagating to neighboring cells (Fig. 7A). In addition to evoking a response in the same cells that responded to the EC25 of CCh, the EC75 concentration of CCh also stimulated more cells to initially respond and greater recruitment of cells for the sustained response (Fig. 7A). The EC25 and EC75 of ATP also produced a similar concentration-dependent activation of cells in the same endothelial preparation (Fig. 7A). Cells that activated Ca2+ signaling in response to the same agonist appeared to form clusters, and the clusters activated by CCh were spatially distinct from those cells activated by ATP (Fig. 7, A and B, and movie S2). The cells activated by each agonist were highly reproducible, and the same pattern of cell activity for each agonist occurred regardless of the sequence of agonist application (fig. S6, A to E).

Fig. 7 Different agonists activate spatially distinct clusters of cells.

(A) Representative composite images of Ca2+ activity in the endothelium in response to the EC25 and EC75 concentrations of CCh and ATP in a single artery. The cells that responded within the first 4 s after agonist application (first responding) and the entire complement of cells responding throughout the 60 s of agonist application (full response) are shown. The boxes outline a region of cells responding to CCh but not to ATP in the same field of endothelium. Images are scaled to show all activated cells. Scale bars, 50 μm. (B) The number of first-responding cells in (A) and their spatial relationships were calculated from Voronoi neighbor analysis and presented as “star maps.” The plot shows the cells that responded to each agonist within the first 4 s of activation (dots) among the total population of cells (gray crosses). Lines connect responding cells to their nearest responding neighbors. (C and D) The number of neighboring cells activated by the EC25 and EC75 of ATP and CCh in the first 4 s (C) and total cells activated in the full response (D). The total number of neighbors (6) was determined for each cell from the Voronoi neighbor analysis (black circles), as well as the number of neighbors activated for each of the conditions (red circles) and the number of neighbors predicted to be activated from a random model (blue circles). The number of neighbors activated for each of the conditions is significantly greater (P < 0.05) from that predicted from a random model of activated cells. The expectation of the number of neighbors activated randomly was determined as the (mean number of neighbors) * (activating fraction) for each agonist. (E) The number of cells activated at the concentrations shown on the x axis as a percentage of the total population of cells in the field (red) and as the percentage of precisely the same cells that are activated by each of the other conditions (blue). (F) The number of first-responding clusters (red, left) and the number of cells in a first-responding cluster (blue, right) that respond at the EC25 and EC75 of CCh and ATP. Data are representative of n = 5 independent experiments from artery preparations from different animals.

To provide a robust statistical description of cell spatial and neighbor relationships, we used a Voronoi tessellation analysis (Fig. 7B). Analysis of Voronoi plots provides a quantitative method for determining the number of neighbors of each cell and the number of coactivating neighboring cells; thus, it is a method for measuring the extent of clustering. The analysis was restricted to the first 4 s of activation, a time long enough to identify the first-responding cells and short enough to minimize the detection of cells activated because of signal propagation (39). This approach was taken in preference to blocking gap junctions because we previously found that the gap junction–blocking compounds carbenoxolone, glycyrrhetinic acid, and GAP-27 each suppressed Ca2+ signals in the endothelium independently of effects on gap junctions (8). The present results show that each cell had, on average, six neighbors (Fig. 7C, black dots), CCh-sensitive cells tended to be close to other CCh-sensitive cells, and ATP-sensitive cells tended to be close to other ATP-sensitive cells (Fig. 7B).

Using Voronoi edge tessellation to define neighbors, we determined the number of coactivating neighbors responding to a given stimulus for every activating cell (Fig. 7, C and D). The null hypothesis was that there is no spatial correlation in the responses that activation of a particular cell is no more likely if its neighbor activates. The expected number of coactivating cells under the null hypothesis was calculated as the mean number of neighbors (6) multiplied by the activating fraction for each agonist. Because this hypothesis can be arrived at by assigning the response versus no response behavior to cells in a spatially random way, we call this the “random model.”

The number of neighboring endothelial cells activated by equivalent (EC25 and EC75) concentrations of CCh and ATP (Fig. 7, C and D, red dots) was significantly greater than predicted by the random model distribution of activated cells (Fig. 7, C and D, blue dots), suggesting that cells that respond to each agonist tend to be neighbors and that cells are arranged in sensing clusters. The difference between the random model and observed response was greater in the first 4 s of activation (Fig. 7C) than in the full-duration dataset (Fig. 7D), presumably because cell-cell propagation increased the number of cells activated after this period and extended signals beyond the initial sensing clusters.

We next analyzed the extent of overlap of clusters that were activated by each agonist. In this analysis, we first identified the cells that responded to the EC25 or EC75 of one agonist and then determined the percentage of those same cells that responded to the corresponding concentration of the second agonist. The EC25 and EC75 concentrations were derived from the concentration-response data over the full duration of the response (Figs. 2D and 4D). From this analysis, it was clear that only a small percentage of cells that responded to CCh (EC25) also responded to ATP (EC25) and, likewise, that only a small percentage of cells that responded to ATP (EC25) also responded to CCh (EC25) (Fig. 7E).

Confirmation that activation of cells in the same population in response to CCh and ATP tended to occur in clusters was derived by identifying interconnected cells in the Voronoi neighbor analysis (Fig. 7F). For this analysis, we measured cluster sizes for CCh and ATP at the EC25 and EC75 concentrations for each agonist. The number of cells in a cluster in the first 4 s of activation increased substantially with agonist concentration for CCh, but the increase was much shallower for ATP (Fig. 7F). This result suggests that the extent of subsequent recruitment of cells after the first 4 s was substantially greater for ATP than for CCh.

The question arises as to why some cells respond preferentially to one agonist compared to the other. At least part of the explanation appears to be the distribution of M3 and P2Y2 receptors. Fluorescence labeling of M3 and P2Y2 receptors in the endothelium shows that the distribution of these two receptors is highly heterogeneous (Fig. 8, A to E). The cells that expressed M3 receptors were spatially distinct from the cells that expressed P2Y2 receptors (Fig. 8, A, B, D, and E). In these experiments, cell boundaries were identified using platelet endothelial cell adhesion molecule 1 (PECAM-1) fluorescence and were used to extract ROIs corresponding to individual endothelial cells.

Fig. 8 Heterogeneous distribution of muscarinic and purinergic receptors.

(A) Representative fluorescence images showing the endothelial cell surface marker PECAM-1 (green), purinergic P2Y2 receptors (red), and muscarinic M3 receptors (blue) in the endothelium. PECAM-1 and P2Y2 receptors were visualized by immunofluorescence, and M3 receptors were visualized using a fluorescently tagged M3 receptor ligand. (B) Expanded view of the boxed region in (A). (C) Negative controls for P2Y2 and M3 fluorescence staining. PECAM-1 labeling shows cell boundaries. The negative control for P2Y2 immunostaining omitted the primary antibody (Ab), and the negative control for M3 staining was generated by labeling with the fluorescent M3 receptor ligand in the presence of an excess of the M3 receptor competitive antagonist 4-DAMP. (D) Summary of fluorescence imaging data showing the amount of M3 staining in the presence and absence of the antagonist 4-DAMP (blue), the amount of labeling in the presence and absence of the P2Y2 primary antibody (red), and the extent of overlap of specific M3 and P2Y2 staining (gray). (E) Summary data showing the percentage of cells with M3 receptor staining (blue), P2Y2 receptor staining (red), and the percentage of cells with both (gray) (n = 3 independent experiments from artery preparations from different animals; *P < 0.05, unpaired t test). Scale bars, 50 μm.

We confirmed the specificity of the P2Y2 antibody used in these experiments (Fig. 8, A to E) using wild-type 1321 N1 cells and 1321 N1 cells stably expressing recombinant human hP2Y2 receptors (1321 N1-hP2Y2; fig. S7, A to D) (40). 1321 N1 cells are a human astrocytoma cell line that do not endogenously produce any of the eight P2Y receptor subtypes or respond to the naturally occurring nucleotide agonist of P2Y receptors (41) and hence were used as a negative control for P2Y2 staining. The P2Y2 antibody labeled 1321 N1-hP2Y2 but not wild-type 1321 N1 cells. M3 receptor staining was also specific because the M3 receptor antagonist 4-DAMP completely blocked binding of the fluorescent ligand we used to label M3 receptors (Fig. 8, C to E).

Heterogeneity in endothelial cell responses to CCh and ATP is not restricted to intracellular Ca2+ release. In another series of experiments, we examined the spatial heterogeneity in NO production using the fluorescent NO indicator DAR-4M (diaminorhodamine-4M) (29). The production of NO largely occurred in distinct cells in response to the EC50 concentration of CCh or ATP for Ca2+ signaling in the same preparation (fig. S8, A to C). In this series of experiments, we first identified the cells that generated NO in response to one agonist and then determined the percentage of precisely the same cells that responded to the second agonist. There was no significant overlap in the cells activated by CCh and ATP to produce NO (fig. S8, B and C).

Together, these results suggest that certain clusters of cells have a higher sensitivity to the muscarinic agonist, whereas different clusters of cells have a higher affinity for purinergic agonists. Thus, the endothelium uses an agonist-specific sensing system that consists of clusters of cells that are sensitive to a particular agonist rather than relying on cells that are equally sensitive to both agonists.

Characteristics of Ca2+signals

The spatiotemporal features of the Ca2+ signals in response to low but equivalent concentrations (EC25) of CCh and ATP evoked in the same tissue preparation differed substantially (Fig. 9, A to C). In single endothelial cells, a low concentration of CCh (EC25) evoked repetitive Ca2+ oscillations on a baseline Ca2+ value that was higher than the Ca2+ baseline before the addition of CCh (Fig. 9, B and C). The mean response from the activated cells (Fig. 9B, green line) showed a slow, steady increase in Ca2+ that remained increased throughout the treatment. In contrast, a low concentration of ATP (EC25) evoked a rapid rise in Ca2+, followed by a decline back to resting values (Fig. 9B). Several cells showed repetitive oscillations in response to ATP. The mean response to the EC25 of ATP (Fig. 9B, red line) was a sharp increase in Ca2+ that subsequently declined toward baseline. When CCh and ATP (each at their respective EC25) were applied together, the cells that responded to the combined treatment were the same cells that responded to the application of each agonist individually (Fig. 9A).

Fig. 9 Signal integration and parallel processing.

(A) Representative composite Ca2+ images showing cells that respond in the first 4 s after application of the EC25 concentrations of CCh (green), ATP (red), and both agonists applied simultaneously (cyan). Images are of the same field of endothelium. Scale bars, 50 μm. (B) Ca2+ responses from all activated cells in the field of endothelium shown in (A). Agonists were present for the duration indicated by the line above each trace. (C) Examples of responses from three separate cells to CCh, ATP, and both agonists applied together. In each panel, traces are from the cells indicated by the numbered white dots in (A). (D) Mean peak responses (black circles) to the EC25 of CCh and ATP separately and when both were applied together. The red line shows the calculated mean of peak response when both agonists were added separately, and the red-shaded region indicates the SEM for this measure. The blue line shows the sum of the peak responses when both agonists were added separately, and the blue-shaded region indicates the SEM for this measure. (E) Mean steady-state responses (black circles) to the EC25 of CCh and ATP separately and when both were applied together. The red line shows the calculated mean of the steady-state response when both agonists were added separately, with the red-shaded region indicating the SEM. The blue line shows the sum of the steady-state responses when both agonists were added separately, with the blue-shaded region indicating the SEM. Data are representative of n = 5 independent experiments from artery preparations from different animals.

When both agonists (CCh and ATP) were present, the Ca2+ responses appeared to be a composite of the amplitudes and temporal features of the signals derived from each agonist when they were applied separately (Fig. 9, B and C). Thus, when compared to the response to CCh alone, cells responding when both agonists were present showed a sharper increase in Ca2+, followed by a sustained phase with a higher amplitude in [Ca2+] (Fig. 9B).

The change in characteristics of the Ca2+ response occurred in individual cells and was not a consequence of the analysis averaging the responses across cells. To show this, we analyzed the responses of three separate cells to CCh alone, ATP alone, and both agonists simultaneously (Fig. 9C). Cell 1 responded to CCh but not to ATP, and although the cell did not exhibit a Ca2+ change in response to ATP alone, the characteristics of the response to CCh were altered when ATP was also present, and there was a faster and larger upstroke (Fig. 9C). Cell 2 responded to ATP but not to CCh, but the presence of CCh altered the response of this cell to ATP, causing a smaller initial peak and a more sustained later Ca2+ change (Fig. 9C). Cell 3 responded to both CCh and ATP (Fig. 9C). Once again, the characteristics of the response of the cell were altered when ATP and CCh were both present.

The combined response that occurred when both agonists were present was neither the mean nor a linear summation of the individual responses. The mean peak in response to CCh and ATP when both agonists were present exceeded the mean response to CCh and ATP when they were applied separately and was less than the summed response (Fig. 9D). The combined steady-state response exceeded both the mean and the summed response to CCh and ATP when applied separately (Fig. 9E). Thus, the intracellular computation that integrates these responses appears to be nonlinear for both the peak and the steady-state response and, in the case of the steady-state response, synergistic, in that the combined response exceeds a linear summation of each input.

Communication among cells

The distinct composite Ca2+ signals that arose when both agonists were present may have occurred because the cells were sensitive to both activators but only sensitive to one of the agonists below the threshold for initiating Ca2+ signaling or because signals from each agonist are communicated between cells. Our experiments provided evidence for the existence of both explanations. To determine whether cell-cell communication occurred, we first identified cells that were largely unresponsive to one activator at each of the EC25 and EC75 concentrations from a field of endothelium (Fig. 10, A to D). In an example highlighted cell (Fig. 10, C and D), the response to CCh at the EC25 or EC75 was small and had little concentration-dependent increase, indicating subthreshold behavior. This same cell responded strongly to ATP (Fig. 10, C and D). The response of this cell to ATP (EC25) was a small Ca2+ peak that declined toward resting values. ATP (EC75) evoked a larger transient Ca2+ rise that then declined to a small, but sustained, increase in baseline. Although the cell responded only weakly to CCh alone, the addition of CCh (EC25) substantially altered the response to ATP (EC25). At the EC25 of ATP, when CCh (EC25) was also present, there was a larger and more sustained increase in Ca2+ with oscillations, compared to when ATP (EC25) alone was present. Because responses to the EC25 and EC75 of CCh were, at best, weak, the modification of the response to ATP (EC25), when the EC25 of CCh was present, is consistent with the cell responding directly to ATP (EC25) but indirectly to CCh through signals from neighboring CCh-sensitive cells.

Fig. 10 Signal integration and communication across cells.

(A) Composite Ca2+ images showing cells that respond in the first 4 s after the application of the EC25 and EC75 concentrations of CCh (green) or ATP (red) and to the EC25 concentrations of CCh and ATP applied simultaneously (cyan). All images are from the same field of endothelium. (B) Ca2+ responses from the cells in (A), including 30 s of baseline measurements before agonist application and 60 s after agonist application. (C) Higher magnification view of the boxed area in (A) of an example cell (yellow outline) showing its responses to low (EC25) and high (EC75) concentrations of CCh and ATP. (D) Ca2+ responses of the individual cell shown in (C). Scale bars, 50 μm. Data are representative of n = 3 independent experiments from artery preparations from different animals.

Some cells were sensitive to both activators. In those cells, we found that the Ca2+ response was distinct for each agonist (Fig. 9C). A low concentration of CCh (EC25) still evoked repetitive Ca2+ oscillations on an increased baseline value, whereas a low concentration of ATP (EC25) evoked a rapid rise in Ca2+, followed by a decline back to resting values. This observation suggests that the distinctive features of the Ca2+ signals evoked by each agonist were not intrinsic features of the cells activated by each agonist (which would mean that separate cells have only one type of response) but were rather a feature of the agonist activation of the cells.

DISCUSSION

A major challenge faced by the endothelium is the need to process an enormous quantity of information held in the complex chemical environment to which the vascular system is exposed. From this information, decisions on physiological outputs are made. The endothelium uses a multitude of receptors to continuously monitor vanishingly small changes in the concentration of extracellular signals (25, 42), each of which provides cues about the ever-changing physiological state. Extracellular signals must be accurately detected and correctly relayed through intracellular Ca2+ signals and other intracellular signaling pathways so that information is not lost. Precisely how the endothelium processes the entire chemical composition to which it is exposed and transduces multiple extracellular signals simultaneously to elicit specific cellular responses is not fully understood. We have shown that the endothelium uses spatially structured arrangements of cells that extract discrete components of the overall information content. These components evoke distinctive Ca2+ and NO signals in specific clusters of cells. The Ca2+ signals are shared among cells. When more than one stimulus is present, computations are carried out on the input signals to generate new composite signals that are not a simple sum of the input signals. The endothelium thus detects signals in a noisy environment by using cells that are not sensitive to all possible available information but only sensitive to a small part of the overall content. These cells then collaborate with network partners to distribute and integrate the information.

The endothelium has a very large functional repertoire and controls virtually every cardiovascular function. Dysfunction of the endothelium underlies nearly all cardiovascular disease. Short-range network properties may determine the endothelium’s sensing abilities and the effectiveness of information transfer. Despite the diverse functional activities of the endothelium, the anatomy and connectivity of the cells and network are largely fixed, with each cell having six neighbors on average. The question of how a large functional signaling repertoire arises from a fixed anatomy and connectivity using predominantly a single communication signal (Ca2+) both within and between cells is unresolved. Most studies on the endothelium provide information derived from large ensemble measurements in intact arteries or cultured cell models in which the behavior of the population is assumed to be uniform. When communication between cells is believed to occur, information arriving at one cell is thought to be conveyed to neighboring endothelial cells without changing the information content, much like signaling through a cable (11, 12). We developed an approach that provides an independent readout of hundreds of individual endothelial cells in intact arteries. The findings show that there is divergence between structural and functional networks. Although the endothelium has a defined, fixed network architecture, responses to different extracellular signals are organized into separate local clusters and interactions within that network ensure efficient message detection and permit multiple simultaneous stimuli to be accommodated.

Our results show that the response throughout the endothelium was not uniform, and various regions selectively responded to different activators. In those regions, endothelial cells that activate Ca2+ signaling in response to a particular agonist were found to be statistically significantly more closely spaced (clustered) than expected if the agonist response was randomly distributed across cells. Different agonists activated different clusters of cells in separate regions of the endothelium. The clustering of cells with responsiveness to particular agonists may offer several advantages. Clustering may provide a coincidence detection system to help improve signal detection (8). It is also tempting to speculate that clustering may allow the uptake and breakdown mechanisms for diffusible messengers, such as NO and prostaglandins, to be overwhelmed, thus allowing increased spillover of signals to nearby cells. Clustering may limit the interference from neighboring cells that are responding to a different stimulus. Perhaps, a single cell responding in isolation may easily be influenced by neighboring cells and have its signal overridden, whereas a cluster of cells responding similarly may be much harder to override. The mechanisms giving rise to the organization of cells into clusters are not yet clear, but perhaps clonal amplification of cells with specific sensitivities occurs during development, there is feedback control of function and receptor abundance based on location, or there is another self-organization process occurring at the cellular level.

Although the endothelium is usually treated as being a homogeneous population of cells, there is evidence of heterogeneity in the sensitivities and responses of endothelial cells within the same region of vasculature. Various receptors have been reported to be heterogeneously distributed in the endothelium (43). The distribution of angiotensin II is irregular in neighboring endothelial cells of femoral artery and aorta, either because of differences in uptake mechanisms or synthesis of angiotensin II within cells (44). Acetylcholine (ACh)–evoked Ca2+ responses are larger at branches in rat thoracic aorta than in nearby nonbranch regions (38), and the reverse is true of histamine-evoked Ca2+ responses (38). The sensitivity to histamine and ACh is not distributed evenly among neighboring cells but arranged in “belts” of high sensitivity that varied by ~100 fold along the flow lines (38). In studies of murine thoracic aorta endothelial cells, although most cells (82%) respond to ATP, large fractions of cells do not respond to ACh, bradykinin, or substance P (45). A mosaic pattern of microdomains of von Willebrand factor–positive and von Willebrand factor–negative endothelial cells exists in the capillaries of many vascular beds and in the aorta (6, 44, 46). Our results also show that there is heterogeneity in the distribution of M3 and P2Y2 receptors, which offers an explanation for the heterogeneous responses of endothelial cells to CCh and ATP stimulation. Single-cell profiling of primary endothelial cells revealed an unexpectedly high amount of heterogeneity and patterning of G protein–coupled receptors in subpopulations of cells (47). Although appreciated to exist, the physiological importance of the various receptor distributions and sensitivities has not been elucidated. The heterogeneity in the distribution of receptors and clustering of sensing cells may be central to the function of the elaborate sensing system operating in the endothelium.

Many (1316, 29), although not all (20), extracellular activators elicit changes in intracellular Ca2+ concentration to evoke specific cell responses. An unresolved question in Ca2+ signaling is how this single ion can selectively control virtually all cellular processes. Our data show that Ca2+ signals evoked by low but functionally equivalent (EC25) concentrations of CCh and ATP in separate cells are distinctive. CCh evoked a small persistent rise in Ca2+, whereas ATP evoked an initial Ca2+ spike, followed by a decline to basal values. Thus, there is an input-output relationship between agonist and signal that is temporally distinctive. When the agonists were applied separately at moderate concentrations (EC25), most cells were sensitive to only one agonist (unimodal cells). However, some cells were sensitive to both agonists (multimodal cells). When the agonists were added separately, multimodal cells generated distinctive signals in response to each agonist. This observation suggests that the distinctive signals evoked by each agonist are a feature of the agonist acting on the cells rather than the cell itself.

The presence of both agonists simultaneously generated a new signal output that was distinct from each of the signals generated when the agonists were present separately. The generation of a new signal suggests that cells perform computations by combining information from each source to produce a new output. At least two mechanisms may give rise to the new distinctive signal. First, in the multimodal cells that have receptors for both activators, a summation or averaging of activation of individual responses may have occurred. Alternatively, unimodal cells may be directly activated by only one agonist but may receive signals from another cell that is receptive to the other agonist. To test for the possibility that cells also received signals from other cells, we examined signals in (the few) cells responded to neither the EC25 nor EC75 of one agonist. In those cells that did not respond to one agonist even at the EC75, there was nonetheless modification of the signals generated in response to the other agonist when both were present at the EC25. It seems likely that cells received signals [Ca2+ or IP3 (inositol 1,4,5-trisphosphate)] from neighboring cells to modify responses when both agonists were present.

Together, these results suggest that different stimuli evoke distinctive signals in endothelial cells, that the input-output relationship for a given cell and agonist is not fixed, and that cells act as computational elements when more than one signal is present. The precise nature of the computation is not clear, but it appears that it is not a simple summation or averaging of each separate signal. In the case of the steady-state response, the computation appears to be nonlinear and expansive in that the combined response exceeds linear summation of each input. The computation is a feature emerging from the collective dynamics of the endothelial network and provides a mechanism for the endothelium to interactively monitor external environments through sensing that is distributed across separate cells. This arrangement may give rise to the diverse function in the endothelium arising from a fixed network structure that communicates using a single signaling molecule (Ca2+).

The cardiovascular system undergoes continuous bombardment from a multitude of chemical instructions providing information on physiological status. Most signals are generated to evoke the continuous, routine, small adjustments to maintain cardiovascular function. As a result, many of the signals are of a very small magnitude and are superimposed on a baseline concentration that is noisy and close in value to the normal peak increase (25). The normal daily concentration changes of circulating hormones, such as parathyroid hormone, epinephrine, angiotensin II, or leptin, are vanishing small and vary from a resting concentration of a few tens of picomolar to a peak increase in concentration in the low hundreds of picomolar (25, 42). The endothelium acts as the interface for the information and must detect the signals and decode an enormous quantity of data held in multiple signals to direct resources to the most behaviorally relevant cardiovascular activity. The endothelium must also quickly redirect resources when faced with rapid physiological changes involving much higher local concentration changes such as those occurring during a bleeding event. The mechanisms by which the endothelium is capable of merging inputs from multiple agonists of varying concentration magnitudes are unknown. The data in the present study suggest that the processing of multiple chemical instructions is not achieved by each cell sensing every instruction but from the separate activity of a collective of endothelial cells with distinct sensitivities. Various endothelial cells encode only certain aspects of the overall information content. Communication of information among connected cells allows the endothelium to merge information and provide a more comprehensive “picture” of physiological status. Together, our observations suggest that sensing and control are distributed among many endothelial cells and that the spatial structuring of sensory cells and temporal encoding of signals are fundamental to endothelial signal responses. As a collective, the endothelium can efficiently process multiple instructions and act as a sensory system that has substantially greater computational power than the capabilities of any single cell.

MATERIALS AND METHODS

Animals

All animal care and experimental procedures were carried out with the approval of the University of Strathclyde Animal Welfare and Ethical Review Board [Schedule 1 procedure; Animals (Scientific Procedures) Act 1986, UK], under UK Home Office regulations. All experiments used freshly isolated rat carotid arteries obtained from male Sprague-Dawley rats (10 to 12 weeks old; 250 to 350 g) euthanized by intraperitoneal overdose of pentobarbital sodium (200 mg kg−1; Pentaject; Merial Animal Health Ltd., UK) or CO2 overdose.

En face artery preparation

Endothelial Ca2+ signaling was imaged in en face carotid artery preparations (31). Arteries were cut open along their longitudinal axis using microscissors and pinned out on a Sylgard block, with the lumen side upward. Endothelial cells were incubated with a PSS containing the fluorescent Ca2+ indicator Cal-520 acetoxymethyl ester (Cal-520/AM; 5 μM), 0.02% Pluronic F-127, and 0.35% dimethyl sulfoxide for 30 min at 37°C. After incubation, the arteries were gently washed with PSS at room temperature and allowed to equilibrate for 30 min. The Sylgard block containing the artery was placed in a custom bath chamber that was fixed to the stage of an inverted fluorescence microscope (Nikon Ti-E). The bottom of the bath chamber was a 0-grade thickness microscope coverslip. Two 200-μm pins were used to raise the Sylgard block from the coverslip so that the endothelium did not contact the glass coverslip, thus permitting solutions to be flowed through the chamber. Laminar flow was provided by a syringe pump that was connected to the chamber through silicone tubing. Solutions were flowed into the bath at a rate of 1.5 ml/min.

Cal-520/AM was excited with a 488-nm excitation light provided by a monochromator (Photon Technology International/Horiba UK Ltd.) and imaged using a back-illuminated electron-multiplying charge-coupled device (EMCCD) camera (13-μm pixel size; iXon3, Andor) through a 40× (oil immersion; numerical aperture of 1.3; Nikon S Fluor) objective lens (48, 49). Fluorescence emission was recorded at 10 Hz. Fluorescence illumination was controlled, and images (16-bit depth) were captured using WinFluor (University of Strathclyde, Glasgow, UK). After each pharmacological activation, the preparation was washed with PSS for 5 min and then allowed to rest for a further 5 min to ensure that a stable baseline was established before subsequent drug applications (fig. S1G).

Cholinergic and purinergic Ca2+signaling and response to shear stress

Endothelial Ca2+ signaling was examined in response to muscarinic and purinergic agonists. Full noncumulative concentration responses were obtained in each en face preparation by applying each of the agonists at various concentrations. The concentration of CCh varied from 1 nM to 30 μM and that of ATP from 100 nM to 1 mM. Only one concentration-response curve was generated per carotid artery preparation.

In experiments designed to analyze the first-responding cells (at EC25 and EC75 concentrations), the agonists were sequentially applied to the same artery preparation, and recordings were made from the same field of endothelium for the EC25 and EC75 concentrations for both agonists. The EC25 and EC75 values were obtained from concentration-response curves showing the “percentage of active cells.” Both agonists were then applied simultaneously, at the EC25 concentration, to the same artery to determine any interaction between the two agonists in the cells activated.

In experiments designed to determine the receptor subtype involved in CCh-evoked Ca2+ signaling, a selective M3 antagonist 4-DAMP (100 nM) was used. To determine whether metabotropic or ionotropic receptors were involved in ATP-evoked Ca2+ signaling, ATP was applied to the endothelium in Ca2+-free PSS. Specific P2Y1 receptor antagonist (MRS2179; 10 μM) and P2Y2 receptor antagonist (ARC118925; 1 μM) were used to discern the metabotropic receptor subtype involved in ATP-evoked Ca2+ signaling.

In experiments examining shear stress, laminar flow (1.5 ml/min; ~2.5 dyne cm−1) was provided by a syringe pump. Ca2+ signals were measured from whole-cell ROIs. The endothelial Ca2+ response to shear stress was compared in the same field to endothelium to the response to ACh (300 nM) added to the perfusion solution.

Automated Ca2+signal analysis

Because of the large number of cells imaged, manual analysis of each individual cell was impractical. Therefore, automated analysis software (WAVE; whole-cell average) was used to extract Ca2+ signals from every cell in the field of view (31). Briefly, (i) all cells within the field of view are identified, and an ROI that encompasses the cell area is created; (ii) the Ca2+ signals from each ROI are extracted as time-dependent fluorescence intensities; (iii) each Ca2+ signal is normalized to its corresponding baseline; (iv) Ca2+ signaling metrics, such as peak amplitudes and oscillation frequency, are rapidly and objectively determined for all cells (ROIs); and (v) Ca2+ signals and summary data are presented in pictorial form. The Ca2+ signal from precisely the same cell can be matched across different experiments.

ROI generation

Intensity thresholding of the raw image does not yield accurate cellular ROIs due to the high density of endothelial cells in intact arteries. Instead, average intensity projections were sharpened using an unsharp masking macro written in ImagePro (50). After sharpening, individual cellular ROIs were obtained by intensity thresholding (in ImagePro) and manually splitting any ROI that overlapped two (or more) cells. To enable a comparison (pairing) of the response of individual cells after various treatments during an experiment, cellular ROIs generated from a single control dataset were applied to all responses obtained in that single experiment. Where slight tissue movement occurred between acquisitions, ROIs were aligned across individual datasets using an automated alignment plug-in in FIJI (50). Each ROI (cell) was assigned an individual number (cell #) for identification throughout and to enable a direct comparison of the same cell in response to different treatments.

Automated whole-cell Ca2+signal extraction

Ca2+ signals were calculated for each individual cell from the Cal-520/AM fluorescence intensity within each ROI. Each signal was expressed as baseline-corrected fluorescence intensity (F/F0), where F is fluorescence at time, t, and F0 is basal fluorescence intensity. F0 was determined for each individual ROI by averaging a 100-frame (10 s) baseline period at the start of each recording. F/F0 traces were smoothed using 11-point (1.1 s), third-order polynomial Savitzky-Golay filter.

Automated whole-cell Ca2+signal analysis

Endothelial Ca2+ signals exhibit oscillatory behavior. The occurrence of these oscillations was identified automatically using the discrete (first) derivative (d(F/F0)/dt) of the Ca2+ signal (F/F0). Discrete derivate signals were calculated by convolving F/F0 traces with the first derivative of a Gaussian kernel in the programming language, Python (8, 51). An increase in F/F0 corresponds to a positive deflection in the discrete derivative, and a decrease in F/F0 corresponds to a negative deflection in the discrete derivative. At the peak of an F/F0 Ca2+ spike, the derivative changes sign from positive to negative. A “zero-crossing detector” was used on the discrete derivative to identify the time at which peaks occurred in the F/F0 data, i.e., when the sign of the rate changed from positive to negative and the rate of change is zero. The zero-crossing detector provided a list of times of “zero crossing,” i.e., the times at which a peak occurs in the data. Therefore, the zero-crossing detector enabled the times of all peaks, with a magnitude greater than 10 times the SD of baseline noise (considered to be the threshold), in the derivate signal to be extracted. Those times were then used to extract conventional measurements (e.g., amplitude) from the corresponding F/F0 data. The detection of a peak from the discrete derivative trace (above threshold) was used to determine whether a cell responded. If a peak was detected in a cell, then the cell was defined as being “active.” Oscillatory Ca2+ signals evoked by the agonists were measured in terms of peak amplitude of the initial Ca2+ rise, the response averaged over 60 s after this peak, and percentage of the total number of cells which responded. The times of occurrence of the first peak in each Ca2+ response were used to generate latency profiles of cellular Ca2+ responses. These latency profiles are presented as histograms with time, t = 0, corresponding to the time of the first detected peak.

Ca2+ responses were recorded over a 5-min period. This allowed for the recording of basal Ca2+ activity (no agonist; ~1 to 2 min) and the response to stimuli (~3 to 4 min).

Cell clustering and neighbor analysis

To examine the spatial relationship of cell responses, a robust means to define spatial relationships between cells was required. For this purpose, we used a modification of a Voronoi tessellation (52, 53). Briefly, the center coordinates of each biological cell’s nucleus form a vertex in a Voronoi tessellation. The tessellation partitions the image space into a series of Voronoi cells. Each Voronoi cell contains all points closer to its vertex than to any other vertex. Each edge in the Voronoi diagram separates two Voronoi cells and represents all points in space equidistant to the coordinates of the two nuclei that define the two Voronoi cells and that are closer to those nuclei than to any other. We used each edge in the tessellation to define the corresponding biological cells to be “Voronoi neighbors.” Voronoi tessellation and subsequent analysis were carried out in the Python language version 2.7.11 using the SciPy (54) and NumPy (55) libraries, versions 0.17.0 and 1.10.4, respectively.

NO imaging and analysis

Arteries were pinned flat in a custom-made bath chamber and preloaded with Cal-520 as described above to identify endothelial cell positions. To measure NO production, arteries were then incubated in Hanks’ buffer solution containing DAR-4/AM, a cell-permeable NO reporting fluorophore (100 μM DAR-4/AM, 0.02% Pluronic F-127, and 100 μM l-arginine), for 30 min at room temperature. After incubation, arteries were gently washed with Hanks’ buffer and allowed to equilibrate for 30 min. Laminar flow was provided by a syringe pump that was connected to the chamber via silicone tubing. Solutions were flowed into the bath at a rate of 1.5 ml/min. DAR-4M AM was excited using 550-nm wide-field epifluorescence illumination provided by a monochomator (Photon Technology International/Horiba UK Ltd., Stanmore, UK) and was visualized using a back-illuminated EMCCD camera (13-μm pixel size; iXon Life 888, Andor) through a 60× (water immersion; numerical aperture of 1.0; Nikon S Fluor) objective lens. Fluorescence emission was recorded at 10 Hz. Fluorescence illumination was controlled, and images (16-bit depth) were captured using μManager software (56). After each pharmacological activation (with CCh or ATP), the preparation was washed with Hanks’ buffer for 10 min and then allowed to rest for a further 10 min to ensure that a stable baseline was established before subsequent drug applications.

Automated analysis was carried out as described above to calculate the NO production for each individual cell. ROIs were generated using the Cal-520 loading, imaged at 488 nm. NO production was recorded over a 5-min period. This allowed for the recording of basal NO activity (no agonist; ~1 to 2 min) and the response to stimuli (~3 to 4 min).

Endothelial immunocytochemistry

Arteries were pinned to a six-well plate containing Sylgard and fixed in 4% paraformaldehyde for 20 min at room temperature. Arteries were washed three times in 100 μM glycine solution, followed by three washes with phosphate-buffered saline (PBS), permeabilized with 0.2% Triton X-100 in PBS for 10 min, and washed again three times with PBS. Arteries were then incubated in antibody buffer solution in PBS for 1 hour at room temperature and then washed three times with antibody wash solution containing 5% bovine serum albumin (BSA). The arteries then were incubated with a rabbit anti-P2Y2 receptor antibody (1:100; sc-20124, Santa Cruz Biotechnology) and goat anti-CD31/PECAM-1 (1:200; AF3628, R&D Systems) in PBS containing 1% BSA and 2% Donkey serum for 12 hours at 4°C. Arteries were washed with antibody wash solution containing 5% BSA three times, followed by incubation with Alexa Fluor 488 donkey anti-rabbit antibody (Invitrogen) and Alexa Fluor 568 donkey anti-goat antibody (Invitrogen) in antibody buffer solution for 1 hour at room temperature. The P2Y2 receptor secondary antibody was an Alexa Fluor 488 donkey anti-rabbit antibody (1:1000) and excited using 488 nm. The CD31/ PECAM-1 secondary antibody was an Alexa Fluor 568 donkey anti-goat secondary antibody (1:1000) and was excited using 568 nm. Excitation light was provided by wide-field epifluorescence illumination via a monochomator (Photon Technology International/Horiba UK Ltd.) and visualized using a back-illuminated EMCCD camera (13-μm pixel size; iXon Life 888, Andor) through a 60× (water immersion; numerical aperture of 1.0; Nikon S Fluor) objective lens. Fluorescence emission was recorded at 10 Hz. Fluorescence illumination was controlled, and images (16-bit depth) were captured using ImageJ (National Institutes of Health, Bethesda, MD, USA). Negative controls were performed in the absence of primary antibody.

Fluorescent ligand binding in intact endothelium

To visualize P2Y2 receptors and M3 receptor distribution, P2Y2 antibody–stained arteries were loaded with a fluorescent muscarinic M3 receptor antagonist (M3-633-AN, 100 nM; Sigma-Aldrich) for 20 min at room temperature. The M3-633-AN was excited at 633 nm and visualized as described above. Negative controls were performed in the presence of the selective M3 receptor antagonist 4-DAMP (10 μM at 20-min incubation before M3-633-AN loading). P2Y2, M3, and PECAM visualization was carried out in precisely the same area of endothelium by changing the excitation wavelength.

Cell boundaries were identified using PECAM-1 fluorescence. From these boundaries, ROIs were manually extracted to identify individual endothelial cells. The area of M3, P2Y2 expression, and the extent of colocalization (M3 and P2Y2 overlap) were then determined within each cell (using the ROIs) and from the full field of view using ImageJ.

P2Y2 receptor antibody validation

Specificity of the antibody for P2Y2 receptors (1:100; sc-20124, Santa Cruz Biotechnology) was confirmed in experiments carried out in cell lines that either (1) completely lack P2Y receptors or (2) express only P2Y2 receptors (fig. S7, A to C). 1321 N1 cells (catalog no. 86030402; The European Collection of Authenticated Cell Cultures; RRID:CVCL_0110) are a human astrocytoma cell line that does not endogenously express any of the eight P2Y receptor subtypes or respond to the naturally occurring nucleotide agonist (41) and were used as a negative control for the antibody. 1321 N1-hP2Y2 (40) were used as a positive control (fig. S7, A to C).

Wild-type 1321 N1 cells and 1321 N1-hP2Y2 (40) were maintained in 5% CO2, 95% O2 in a humidified incubator at 37°C, in Dulbecco’s modified Eagles media (catalog no. 21969-035; Life Technologies, Paisley, UK), supplemented with 10% fetal calf serum, 1% nonessential amino acids, 1% penicillin (10,000 units/ml), and streptomycin (10 mg/ml).

The wild-type 1321 N1 cells and 1321 N1-hP2Y2 were fixed in 4% paraformaldehyde for 20 min at room temperature and washed three times in 100 μM glycine solution, followed by three washes with PBS. Cells were then incubated in antibody buffer solution in PBS for 1 hour at room temperature and then washed three times with antibody wash solution containing 5% BSA. The cells were then incubated with a rabbit anti-P2Y2 receptor antibody (1:100; sc-20124; Santa Cruz Biotechnology, Dallas, TX, USA) in PBS containing 1% BSA and 2% Donkey serum for 2 hours at 4°C. Cells were washed three times with antibody wash solution containing 5% BSA, followed by incubation with Alexa Fluor 488 donkey anti-rabbit antibody (Invitrogen) in antibody buffer solution for 1 hour at room temperature. The P2Y2 receptor antibody was excited using 488-nm wide-field epifluorescence illumination provided by a light-emitting diode (CoolLED pE-300ultra, CoolLED Ltd.) and visualized using a back-illuminated EMCCD camera (13-μm pixel size; iXon Life 888, Andor) through a 40× (oil immersion; numerical aperture of 1.3; Nikon S Fluor) objective lens. Fluorescence emission was recorded at 10 Hz. Fluorescence illumination was controlled, and images (16-bit depth) were captured using ImageJ (National Institutes of Health, Bethesda, MD, USA). Cells were incubated in PBS containing 4′,6-diamidino-2-phenylindole (DAPI; 0.5 mg/ml) to visualize the nucleus of endothelial cells. DAPI was excited at 365 nm and visualized as described above.

P2Y2 receptor distribution was quantified as fluorescence intensity levels. Because there was an almost complete absence of staining in 1321 N1 cells (which lack P2Y receptors), full-field fluorescence was measured. Cell nuclei counts using DAPI staining ensured comparable cell number in wild-type and P2Y2-expressing cells.

Pressure myography

The left and right common carotid arteries were exposed by sharp dissection, and both arteries were extracted and placed in Hanks’ buffer solution. Under a stereomicroscope, arteries were cleaned of connective tissue and visually checked for the presence of side branches. Arteries without side branches were mounted onto two blunted, sterile 25 G by 5/8″ BD Microlance Needles (outer diameter of 0.51 mm) using silk suture thread, in a self-heated single-vessel chamber (CH-1-AU-SH, Living Systems Instrumentation). Arteries were gently heated to 37°C and allowed to equilibrate for 30 min. The proximal end of each BD Microlance Needle was connected to a separate syringe containing Hanks’ buffer solution. The height of two syringes was adjusted to alter the transmural pressure of the arteries. The pressure was increased in 10-mmHg increments, with 10-min equilibration allowed before subsequent increases in pressure. When a final pressure of 110 mmHg was attained, arteries were allowed to equilibrate for 30 min. During equilibration, arteries were tested for leaks (which may indicate branches in the vessel wall or insufficiently tied sutures) by stopping the flow of solution from the syringes to the arteries by switching the valves connecting the syringes to the BD Microlance Needles. This allowed the pressure within the arteries to remain constant. Arteries that showed a reduction in pressure were discarded. Hanks’ buffer was circulated in the chamber that contained the arteries at 10 ml/min using a peristaltic pump from a reservoir (heated to 37°C).

To determine the vasorelaxant effects of CCh and ATP, arteries were first contracted using phenylephrine (100 μM), applied to the outside of the artery by adding the agonist to the reservoir and allowing it to circulate. Arteries were then allowed to equilibrate to ensure that a stable contraction was achieved before subsequent drug applications. CCh or ATP was applied to the arteries intraluminally. To do this, the height of one syringe was adjusted to allow gravitational flow (0.5 ml/min) of solution through the lumen of the artery. CCh (100 μM) was applied by adding the drug to the higher of the two syringes that perfused the artery lumen. Arteries were then allowed to equilibrate for 20 min (10 min for the drug to reach the lumen; 10 min of activation). After activation, the solution in the syringe was removed and replaced with Hanks’ buffer. Arteries were washed intraluminally with Hanks’ buffer until arteries returned to a contracted state. ATP (100 μM) was subsequently applied to the same artery, as described above.

In experiments designed to determine endothelial Ca2+ dependence of CCh- and ATP-mediated vasodilation, BAPTA-AM (30 μM), a cell-permeable selective Ca2+ chelator, was applied intraluminally as described above. Preconstricted arteries were allowed to equilibrate in BAPTA-AM for 30 min before the addition of CCh (100 μM) and ATP (100 μM). The effects of the Ca2+ chelator (BAPTA) on CCh- and ATP-evoked relaxation were compared to controls (absence of BAPTA) carried out in the same artery.

Artery images were recorded using a complementary metal-oxide semiconductor camera (5.2-μm pixel size; DCC1545; ThorLabs Inc., Newton, NJ, USA) through a 2.5× (air; numerical aperture of 0.06; Zeiss A-Plan) objective lens. Images (8-bit depth) were captured using μManager software (57).

Reagents and solutions

Cal-520/AM was obtained from Abcam. Pluronic F-127 was obtained from Invitrogen. AR-C 118925XX and MRS 2179 were obtained from Tocris Bioscience. P2Y2 antibody (H-70) was obtained from Santa Cruz Biotechnology. CD31/PECAM-1 antibody was obtained from R&D Systems. All other drugs and chemicals were obtained from Sigma-Aldrich. PSS contains 145 mM NaCl, 4.7 mM KCl, 2.0 mM Mops, 1.2 mM NaH2PO4, 5.0 mM glucose, 0.02 mM EDTA, 1.17 mM MgCl2, and 2.0 mM CaCl2, adjusted to pH 7.4 with NaOH. Ca2+-Free PSS contains 145 mM NaCl, 4.7 mM KCl, 2.0 mM Mops, 1.2 mM NaH2PO4, 5.0 mM glucose, 0.02 mM EDTA, 1.0 mM EGTA, and 2.34 mM MgCl2, adjusted to pH 7.4 with NaOH. Hanks’ buffer contains 125 mM NaCl, 5.4 mM KCl, 4.2 mM NaHCO3, 0.4 mM KH2PO4, 0.3 mM NaH2PO4, 10 mM Hepes, 10.0 mM glucose, 1.17 mM MgCl2, and 2.0 mM CaCl2, adjusted to pH 7.4 with NaOH. Antibody buffer solution contains 0.05% Triton X-100, 0.02% sodium azide, and 5% SSC (20×). Antibody wash solution contains 5% SSC (20×), 0.05% Triton X-100, and 5% BSA. All solutions were freshly prepared daily.

Statistics

The data are presented as means ± SEM values; n refers to the number of animals. For comparison of two groups, a Student’s t test (paired data) or unpaired t test (unpaired data) was performed. All statistical analyses were performed using GraphPad Prism, version 6.0 (GraphPad Software). P < 0.05 was considered statistically significant.

SUPPLEMENTARY MATERIALS

www.sciencesignaling.org/cgi/content/full/11/561/eaar4411/DC1

Fig. S1. CCh-evoked Ca2+ signaling in the endothelium.

Fig. S2. Flow-induced endothelial Ca2+ signaling.

Fig. S3. Ca2+ dependence of CCh- and ATP-evoked endothelial relaxation.

Fig. S4. Concentration dependence of the CCh-evoked endothelial Ca2+ response.

Fig. S5. Concentration dependence of the ATP-evoked endothelial Ca2+ response.

Fig. S6. The sequence of agonist addition does not alter the pattern of cell activation.

Fig. S7. Specificity of the antibody recognizing the P2Y2 receptor.

Fig. S8. Heterogeneity in NO production.

Movie S1. Regional Ca2+ signals evoked by 100 nM CCh.

Movie S2. Spatially distinct cell clusters activated by the EC25 concentrations of CCh and ATP.

REFERENCES AND NOTES

Acknowledgments: We are grateful for the excellent technical support from M. MacDonald. Funding: We acknowledged support from the Wellcome Trust (202924/Z/16/Z) and the British Heart Foundation (PG/16/54/32230 and PG16/82/32439). C.W. is supported by a Sir Henry Wellcome Postdoctoral Research Fellowship (204682/Z/16/Z). Author contributions: Conceptualization: J.G.M., C.W., and M.D.L.; methodology: M.D.L., C.W., C.D.S., J.M.G., and J.G.M.; investigation: M.D.L., C.K., and C.W.; writing (original draft): M.D.L. and J.G.M.; writing (review and editing): J.G.M., M.D.L., C.W., C.D.S., C.K., and J.M.G.; funding acquisition: J.G.M., C.W., C.D.S., and J.M.G.; resources: J.G.M., C.W., C.D.S., and J.M.G.; supervision: J.G.M. and C.W. Competing interests: The authors declare that they have no competing interests. Data and materials availability: All data needed to evaluate the conclusions in the paper are present in the paper and/or the Supplementary Materials.
View Abstract

Stay Connected to Science Signaling

Navigate This Article