Localized expression of an Ins(1,4,5)P₃ receptor at the myoendothelial junction selectively regulates heterocellular Ca²⁺ communication

The functional integration of endothelial cells (ECs) and vascular smooth-muscle cells (VSMCs) within arterioles is important for several physiological processes, including the control of blood flow and response to vascular wounding (Figueroa et al., 2004; Figueroa et al., 2006; Michel et al., 1995). The integration of ECs and VSMCs can occur via the release of paracrine factors [e.g. nitric oxide (NO)] or through direct cell-cell contact via gap junctions at sites termed myoendothelial junctions (MEJs) (e.g. Sandow and Hill, 2000).

Gap junctions are dodecameric channels composed of two hexameric hemichannels that allow movement of current and solutes (<1000 Da) directly from the cytoplasm of one cell to the cytoplasm of an adjacent cell (Figueroa et al., 2004). Connexin (Cx) proteins compose the hemichannels and are found in over 20 different isoforms, each conferring electrical or solute selectivity on the hemichannel or gap junction (Saez et al., 2003). Although hemichannels composed of a single connexin isoform (homomeric) have been studied extensively, the mixing of connexin isoforms into hemichannels (heteromeric) is probably a common occurrence because cells express multiple connexin isoforms (Koval, 2006). The mixing of connexin isoforms in hemichannels and in gap junctions probably has important implications regarding the regulation of solute movement (Locke et al., 2004), which might be especially important in MEJs, where the presence of multiple connexin isoforms might lead to the formation of heterotypic gap junctions. For example, in rat mesenteric arteries, both Cx37 (also known as Cxa4 and Gja4) and Cx40 (also known as Cxa5 and Gja5) have been demonstrated to be present between ECs and VSMCs, and, in mouse cremaster arterioles, Cx40 and Cx43 (also known as Cxa1 and Gja1) predominate (Isakson et al., 2008). Because certain heterotypic gap junctions have been shown to be electrically rectifiable (Kreuzberg et al., 2005) and some heteromeric hemichannels have been shown to be less permeable than homomeric channels to the second messenger inositol (1,4,5)-trisphosphate [Ins(1,4,5)P₃] (Ayad et al., 2006; Locke et al., 2004), it is possible that heterotypic gap junctions at the MEJ are capable of selectively regulating the movement of solutes.

Several laboratories have now demonstrated that stimulation of VSMCs with phenylephrine (PE) via α1_D-adrenoceptors causes an increase in VSMC intracellular Ca²⁺ concentration ([Ca²⁺]_i), with a subsequent increase in EC [Ca²⁺]_i (Dora et al., 1997; Isakson et al., 2007; Jackson et al., 2008; Kansui et al., 2008; Lamboley et al., 2005). This intercellular Ca²⁺ communication is probably due to the movement of Ins(1,4,5)P₃ from VSMCs, through gap junctions at the MEJ, to ECs (Isakson et al., 2007; Kansui et al., 2008; Lamboley et al., 2005). However, stimulation of ECs with agonists that are known to generate Ins(1,4,5)P₃ produces no observable increase in Ins(1,4,5)P₃ within VSMCs (e.g. de Wit et al., 2006), suggesting unidirectional movement of Ins(1,4,5)P₃ through the gap junctions at the MEJ. An attractive hypothesis is that gap junctions at the MEJ provide directionality for the movement of Ins(1,4,5)P₃.

The half-life of Ins(1,4,5)P₃ is approximately 60 seconds (Sims and Allbritton, 1998) before being metabolized [e.g. by type-I 5-phosphatases (Gaspers and Thomas, 2005; Rottingen and Iversen, 2000; Safrany et al., 1994; Speed et al., 1999)]. However, numerous reports have demonstrated that Ins(1,4,5)P₃ is sufficiently stable to survive movement through gap junctions and thus increase [Ca²⁺]_i in neighboring cells (e.g. Boitano et al., 1992; Carter et al., 1996). The secondary response to Ins(1,4,5)P₃ in neighboring cells is dependant on Ins(1,4,5)P₃ binding to Ins(1,4,5)P₃-receptors [Ins(1,4,5)P₃-Rs] on the endoplasmic reticulum (ER) of the neighboring cell and a subsequent release of Ca²⁺. Three isoforms of Ins(1,4,5)P₃-R exist [Ins(1,4,5)P₃-R1 (also known as ITPR1), Ins(1,4,5)P₃-R2 (also known as ITPR2) and Ins(1,4,5)P₃-R3 (also known as ITPR3)], each with distinct Ca²⁺-release dynamics (Patel et al., 1999) and isoform localization within cellular domains (Colosetti et al., 2003). We previously demonstrated that ER extends into the MEJ both in vivo and in vitro (Isakson et al., 2007), so it is possible that Ins(1,4,5)P₃-Rs reside close to the gap junctions at the EC-VSMC interface, providing a pool of releasable Ca²⁺ for Ins(1,4,5)P₃ moving from VSMCs to ECs through gap junctions at the MEJ.

In this paper, we test two hypotheses: (1) that gap junctions impose directionality on Ins(1,4,5)P₃ and (2) localization of Ins(1,4,5)P₃-Rs at the MEJ is responsible for the observed unidirectional Ins(1,4,5)P₃ response. We test these hypotheses using a vascular-cell co-culture (VCCC) for Ca²⁺ measurements in conjunction with selective application of short interference (si)RNAs. Our results demonstrate a novel heterocellular signaling mechanism implicating the MEJ as a unique signaling domain in the vasculature.

Phenotypically distinct cremasteric EC and VSMC (supplementary material Figs S1 and S2) were assembled in a VCCC (Isakson and Duling, 2005) with ECs retaining a microvasculature bio-marker (supplementary material Fig. S3) (King et al., 2004). Phenotypic confirmation demonstrated an EC monolayer on the top of the Transwell (Fig. 1A) and a VSMC monolayer on the bottom of the Transwell (Fig. 1B), with each cell type growing F-actin cellular extensions to create in vitro MEJs. This therefore represents a similar cell-type configuration to an arteriole, in which MEJs are numerous.

Using the cremasteric VCCC, we examined connexin expression. Neither Cx37 or Cx45 were detected at the in vitro MEJ, whereas Cx40 and Cx43 were found in the cellular extensions on the EC side of the MEJ and Cx43 was found in the cellular extensions on the VSMC side of the MEJ (Fig. 1C-F); this arrangement corresponded to mouse cremasteric arterioles in vivo (Isakson et al., 2008). We next examined Cx40 and Cx43 in the cremasteric VCCC in control conditions, in conditions in which ECs had Cx43 siRNA applied and in conditions in which ECs lacked Cx40 and could not demonstrate any re-arrangement of the connexins (Fig. 1G). To determine whether these different connexin organizations resulted in selective gap junctions at the MEJ, we examined Cy3 coupling from ECs to VSMCs and found that, in controls or conditions in which Cx43 siRNA was applied to ECs, heterotypic gap junctions organized at the MEJ. This was indicated by the presence of biocytin, but not Cy3 dye transfer. By contrast, when Cx40 was deleted from the ECs, Cx43 homotypic gap junctions organized at the MEJ and allowed Cy3 movement to the VSMCs (supplementary material Fig. S4) (e.g. Cottrell et al., 2002; Isakson and Duling, 2005).

On the basis of the selectivity of the gap junctions, we examined whether this could affect Ca²⁺ signaling. In control conditions, cremasteric ECs and VSMCs were able to communicate an increase in [Ca²⁺]_i to unstimulated cells (Fig. 2A,B). When the VCCCs were plated with Cx40^–/– ECs, stimulation of ECs or VSMCs did not result in any changes in intercellular Ca²⁺ communication when compared with controls (Fig. 2C,D). Furthermore, there were no changes in EC Cx43 knockdown when compared with controls (Fig. 2E,F). The only time that intercellular Ca²⁺ communication was altered was when connexins composing the gap junctions at the MEJ were eliminated: Cx40 and Cx43 in ECs (Fig. 2G,H) or Cx43 in VSMCs (Fig. 2I,J). These results are consistent with the inhibition of intercellular Ca²⁺ communication when 18 α-glycyrrhetinic acid (18 α-GA) was added to the VCCC (Fig. 2K,L). On the basis of these observations, altering connexin composition of the gap junctions at the MEJ does not alter heterocellular Ca²⁺ communication.

Because Ins(1,4,5)P₃ has been shown to act unidirectionally from VSMCs to ECs in vivo and in our culture model (Isakson et al., 2007; Kansui et al., 2008; Lamboley et al., 2005), we surmised that it was possible that gross changes in intercellular Ca²⁺ communication mask differences in second-messenger movement when the connexin composition at the MEJ is altered. We therefore loaded ECs with BAPTA and stimulated them with ATP so that only the Ins(1,4,5)P₃ response in VSMCs would be visible if it was present (Isakson et al., 2007). Regardless of the connexin composition at the MEJ, there were no conditions that could demonstrate Ins(1,4,5)P₃ movement from ECs to VSMCs (Fig. 3A-C). Next, we loaded VSMCs with BAPTA, stimulated them with PE and found that no changes in the connexin composition at the MEJ could alter the Ins(1,4,5)P₃-induced increase in [Ca²⁺]_i in ECs (Fig. 3D-F), except when the Ins(1,4,5)P₃-Rs in the ECs were blocked with Xestospongin C (XPC) (Fig. 3G-I). Thus, altering connexin composition of the gap junctions at the MEJ does not alter the apparent unidirectional movement of Ins(1,4,5)P₃ from VSMCs to ECs.

An alternative hypothesis to gap junctions imposing directionality on Ins(1,4,5)P₃ signaling is selective localization of Ins(1,4,5)P₃-Rs on the MEJ. Using immunocytochemistry and immunoblots, we probed for each Ins(1,4,5)P₃-R isoform on transverse sections of mouse cremaster arterioles and from freshly isolated cremasteric ECs and VSMCs, and found expression of each isoform within both ECs and VSMCs (Fig. 4A-F). However, only Ins(1,4,5)P₃-R1 was apparent on actin bridges between ECs and VSMCs (Fig. 4G). This was confirmed in transmission electron microscopy (TEM) sections probed for Ins(1,4,5)P₃-R1 (Fig. 4H). Interestingly, Ins(1,4,5)P₃-R1 appeared to be selectively localized to the EC side of the MEJ (Fig. 4H).

Because of the localization of Ins(1,4,5)P₃-R1 at the MEJ in vivo, we attempted to remove the receptor from the ECs in the VCCC by applying Ins(1,4,5)P₃-R1 siRNA to the EC monolayer. The siRNA did not alter Ins(1,4,5)P₃-R1 expression in VSMCs, nor expression of the other Ins(1,4,5)P₃-R isoforms in either cell type (Fig. 5A,B). Using immunocytochemistry, we visualized Ins(1,4,5)P₃-R1 in control conditions or in conditions in which the Ins(1,4,5)P₃-R1 siRNA was applied to the ECs, and demonstrated selective knockdown of the receptor within the EC (Fig. 5C,D). Although both Ins(1,4,5)P₃-R2 and Ins(1,4,5)P₃-R3 were in EC and VSMC monolayers, neither isoform was detected within the EC or VSMC cellular extensions at the MEJ (Fig. 5E,F). Similar to that found in vivo, using the VCCC we localized Ins(1,4,5)P₃-R1 to the EC side of the MEJ; we were capable of selectively downregulating Ins(1,4,5)P₃-R1 expression.

To test whether Ins(1,4,5)P₃-R1 localization influences the directionality of Ins(1,4,5)P₃ signaling from VSMCs to ECs, a VCCC was prepared with BAPTA in VSMCs and Ins(1,4,5)P₃-R1 siRNA in ECs. Under these conditions, stimulation of VSMCs was unable to induce an increase in EC [Ca²⁺]_i, demonstrating that directionality of Ins(1,4,5)P₃ had been lost (Fig. 6A). Knockdown of Ins(1,4,5)P₃-R2 or Ins(1,4,5)P₃-R3 within ECs (Fig. 6B) did not inhibit the Ins(1,4,5)P₃-mediated signal (Fig. 6C,D). Therefore, localization of Ins(1,4,5)P₃-R1 at the MEJ might play a major role in mediating the EC [Ca²⁺]_i response after VSMC stimulation.

Based on the above observations, it was hypothesized that, without Ins(1,4,5)P₃-R1 on the EC side of the MEJ, the Ins(1,4,5)P₃ must be metabolized, or it would be capable of activating Ins(1,4,5)P₃-R2 and/or Ins(1,4,5)P₃-R3 in the EC monolayer. To determine whether Ins(1,4,5)P₃ is metabolized by 5-phosphatase within the MEJ before it reaches the monolayer, ECs were pre-loaded with Ins(1,4,5)P₃-R1 siRNA and 5-phosphatase inhibitor (5-PI). When the BAPTA-loaded VSMCs were stimulated, an increase in EC [Ca²⁺]_i was evident (Fig. 7A), which was inhibited with XPC (Fig. 7B). This demonstrated that degradation of Ins(1,4,5)P₃ might be an important part of the heterocellular signaling process.

If this is the case, we hypothesize that the lack of an identifiable Ins(1,4,5)P₃-R isoform on the VSMC side of the MEJ might be preventing Ins(1,4,5)P₃ from activating Ins(1,4,5)P₃-R isoforms in the VSMC. To test this idea, VSMCs were loaded with 5-PI, and BAPTA-loaded EC were stimulated. In this instance, an increase in VSMC [Ca²⁺]_i was evident (Fig. 7C), which was inhibited by XPC (Fig. 7D). Therefore, it appears that degradation of Ins(1,4,5)P₃ in the VSMCs, after it moved through gap junctions at the MEJ from ECs, might be the reason that no apparent Ins(1,4,5)P₃-induced increase in VSMC [Ca²⁺]_i is observed after EC stimulation.

So as to better understand the localization of Ins(1,4,5)P₃ degradation, we stained transverse sections of the VCCC for Ins(1,4,5)P₃-R1 and 5-phosphatase (Fig. 7E). Line scans through the pores of the Transwell demonstrate 5-phosphatase found through the pore, but to a lesser degree where the Ins(1,4,5)P₃-R1 was localized on the EC side of the MEJ (Fig. 7F). When taken together, these data indicate that Ins(1,4,5)P₃ is bidirectional across gap junctions at the MEJ, but cellular localization of the Ins(1,4,5)P₃-R on the EC side of the MEJ, and possibly increased expression of 5-phosphatase on the VSMC side of the MEJ, determines the heterocellular [Ca²⁺]_i response.

In this study, we put forth two hypotheses to explain the observation that Ins(1,4,5)P₃ can move across gap junctions at the MEJ from VSMCs to ECs, but not from ECs to VSMCs: (1) that gap junctions themselves impose directionality on Ins(1,4,5)P₃ (e.g. the gap junctions are rectifiable) and (2) that localization of an Ins(1,4,5)P₃-R at the MEJ is responsible for the bidirectional Ins(1,4,5)P₃ response. Our data support the second hypothesis, that Ins(1,4,5)P₃ is bidirectional through gap junctions at the MEJ and that the directionality is probably due to the proximity of the Ins(1,4,5)P₃-R on the EC side of the MEJ. We came to this determination by examining the localization of Ins(1,4,5)P₃-R and the functional effect of Ins(1,4,5)P₃-R deletion on Ca²⁺ communication, as compared with the functional effect of specific connexin-isoform deletion or knockdown. Our experiments in relation to each hypothesis are discussed below.

It is generally assumed that coordination of chemical signals between ECs and VSMCs regulates arteriolar reactivity. Although multiple paracrine factors (e.g. NO) certainly play a major part in controlling arteriolar reactivity, solid evidence has emerged that gap junctions linking the two cell types at the MEJ might also have a role in the coordination of arteriolar activity [e.g. VSMC-to-EC communication and EDHF (Haddock et al., 2006; Kansui et al., 2008)]. Both Cx37 and Cx40 have been reported to be the connexins composing the gap junctions at the MEJ in some vascular beds (Sandow et al., 2006), and Cx40 and Cx43 in others (Isakson et al., 2008), with the common link being Cx40 that is almost exclusively derived from the ECs. Therefore, when Cx40^–/– mice exhibited hypertension (de Wit et al., 2003), and EC-specific Tie-2/Cre Cx43^–/– mice exhibited hypotension (Liao and Duling, 2000), an attractive hypothesis was that the observed alterations in blood pressure were due to alterations in solute movement across the gap junctions at the MEJ, causing impaired communication between ECs and VSMCs. Because Ins(1,4,5)P₃ has been demonstrated to move through gap junctions (Boitano et al., 1992) and to move between VSMCs and ECs (Isakson et al., 2007; Lamboley et al., 2005), we focused on this solute. Our data demonstrate that the deletion of Cx40 from the ECs had no effect on gross changes in intercellular Ca²⁺ communication or on the movement of Ins(1,4,5)P₃ from VSMCs to ECs (Figs 2 and 3). There was also no change when Cx43 was knocked down from the ECs (Figs 2 and 3). We interpret this data to mean that heterocellular communication between ECs and VSMCs is not due to a rectifiable gap junction. Recent evidence now indicates that Cx40 is also found between renin-secreting cells in the kidneys and the Cx40 genetic deletion results in increased renin secretion (Wagner et al., 2007). It is possible that this is an explanation for the hypertension observed in Cx40^–/– mice. Based on our evidence, impaired EC and VSMC communication due to Cx40 or Cx43 deletion at the MEJ is not the cause for the observed changes in blood pressure; however, impairments in homocellular EC or VSMC communication after connexin deletion cannot be discounted (e.g. Figueroa et al., 2003).

Although the data presented herein indicates that connexin composition of the gap junction at the MEJ does not influence second-messenger directionality, there are additional ways in which connexins may regulate second-messenger movement. For example, we cannot discount alterations in the rate of Ins(1,4,5)P₃ or Ca²⁺ flux across the gap junctions that are outside of our temporal detection limit, or other untested solutes that might be more heavily influenced by connexin composition (e.g. K⁺ or cAMP). In addition, we have previously demonstrated in mouse cremaster arterioles that Cx43 appears capable of being phosphorylated at the serine-368 residue (Cx43-S368) on actin bridges between ECs and VSMCs [presumably the MEJ (Isakson et al., 2008)]. In the cremasteric VCCC, there is no detectable Cx43-S368 phosphorylation (data not shown), which is probably due to cell-culture conditions and so this variable was not tested in the present study. Cx43-S368 phosphorylation appears to be capable of altering gap-junction permeability, without closing the channel completely (Lampe et al., 2000). It is therefore possible that, if connexin phosphorylation occurs at the MEJ, another layer of control over second-messenger movement could be present.

A pool of releasable Ca²⁺ in the form of ER extends down into the MEJ from the ECs (Isakson et al., 2007). In Fig. 4, we present evidence that Ins(1,4,5)P₃-R1 is on the EC side of the MEJ, both by identification on actin bridges and by immunohistochemistry on TEM sections. The distribution pattern of Ins(1,4,5)P₃-Rs was replicated in the VCCC. The polarized localization of Ins(1,4,5)P₃-Rs has also been demonstrated by Colosetti et al., who showed that Ins(1,4,5)P₃-R1–EGFP [and Ins(1,4,5)P₃-R3–EGFP], but not Ins(1,4,5)P₃-R2–EGFP, localized to the tight-junction region in MCDK cells when cellular polarity was achieved (Colosetti et al., 2003). It is not clear exactly why only Ins(1,4,5)P₃-R1 was found at the MEJ, although it might be due to the differential properties of each Ins(1,4,5)P₃-R. For example, mice with Ins(1,4,5)P₃-R1 deletion die of ataxia (Matsumoto and Nagata, 1999), whereas Ins(1,4,5)P₃-R2- and Ins(1,4,5)P₃-R3-deleted mice do not exhibit a vascular phenotype (Futatsugi et al., 2005). In addition, Ca²⁺ binding to Ins(1,4,5)P₃-R1 has been shown to have a high affinity and serves to control both the opening and closing of the channel (Bosanac et al., 2004). It is possible then that the Ca²⁺ we previously demonstrated to move from VSMCs to ECs serves to modulate the Ins(1,4,5)P₃-R (Isakson et al., 2007). It is clear that more research into the biophysics of the Ins(1,4,5)P₃-R at the MEJ is required.

Ins(1,4,5)P₃ is known to be discreetly activated in signaling domains within the cytoplasm and near the plasma membrane of cells to control Ca²⁺ signaling events (Delmas et al., 2002; Delmas and Brown, 2002). We expand on this by demonstrating functional relevance of the polarized placement of Ins(1,4,5)P₃-R1 at the MEJ. Using Ins(1,4,5)P₃-R1 siRNA, we knocked down the protein from the MEJs of the ECs, which eliminated the VSMC-induced Ins(1,4,5)P₃ response in the ECs (Figs 5 and 6). However, knockdown of Ins(1,4,5)P₃-R2 or -R3 had no effect on Ca²⁺ communication from VSMCs to ECs (Fig. 6). In mesenteric arteries, recent work has shown that an Ins(1,4,5)P₃-mediated Ca²⁺ event originates from holes in the internal elastic lamina [presumably MEJs (Kansui et al., 2008)], lending strong correlative evidence to our hypothesis that selective localization of Ins(1,4,5)P₃-R on the EC side of the MEJ is responsible for initiating increases in [Ca²⁺]_i in ECs after VSMC stimulation.

The enzyme 5-phosphatase causes degradation of Ins(1,4,5)P₃ to Ins(1,4,5)P₂, significantly reducing its ability to bind to and activate Ins(1,4,5)P₃-Rs (Sims and Allbritton, 1998). The addition of 5-PI to ECs with Ins(1,4,5)P₃-R1 siRNA `rescued' the Ca<sup>2+</sup> response in ECs after VSMC stimulation (Fig. 7). Presumably, this was accomplished by preventing Ins(1,4,5)P<sub>3</sub> degradation in ECs after crossing the gap junctions at the MEJ from VSMCs. The Ins(1,4,5)P<sub>3</sub> could then reach the EC monolayer and activate either Ins(1,4,5)P<sub>3</sub>-R2 or Ins(1,4,5)P<sub>3</sub>-R3 (Fig. 7). This `rescue' was inhibited by XPC in ECs, indicating that it was due to activation of the EC Ins(1,4,5)P₃-Rs after VSMC stimulation. In control conditions, VSMCs did not express Ins(1,4,5)P₃-Rs at the MEJ (Fig. 4). For this reason, we tested whether 5-PI could `rescue' a VSMC response after EC stimulation. As demonstrated in Fig. 7, the addition of 5-PI enabled the VSMCs to respond to ECs with an Ins(1,4,5)P₃-mediated response (as noted by the inhibition of the Ca²⁺ response with XPC). Based on these experiments, we hypothesize that Ins(1,4,5)P₃ is bidirectional across gap junctions at the MEJ. This concept was strengthened by the apparent overexpression of 5-phosphatase in the VSMC extensions as compared with the EC extensions of the VCCC (Fig. 7), which we interpreted as a method by which VSMCs ensure that Ins(1,4,5)P₃ is metabolized. Therefore, the location of the Ins(1,4,5)P₃-R on the EC side of the MEJ enables the EC to respond to Ins(1,4,5)P₃, whereas the VSMC, devoid of Ins(1,4,5)P₃-Rs at the MEJ, cannot.

Our results are not without concerns. For example, the length of the cellular extensions composing the in vitro MEJ are greater than that observed at the MEJ in vivo, which might have had a role in exaggerating the responses that we observed. In addition, the temporal limits by our microscope might have not been fast enough to detect rapid Ca²⁺ signaling events in Figs 2 and 3 relating to the alterations in connexin composition. Lastly, it is unclear why 5-phosphatase appears to be slightly more expressed on the VSMC side of the MEJ, which we predict would actually hinder Ins(1,4,5)P₃ moving from the VSMCs to ECs. This implies that Ins(1,4,5)P₃ in VSMCs should be activated near the MEJ to avoid degradation. However, with the localization of the Ins(1,4,5)P₃-R and reduced 5-phosphatase on the EC side of the MEJ, this would ensure that any Ins(1,4,5)P₃ arising from the VSMCs through the gap junction at the MEJ would have the capability to rapidly induce Ca²⁺ release, and so large amounts of Ins(1,4,5)P₃ are not necessarily required. Indeed, it is possible that signaling from VSMCs to ECs is influenced by the amount of Ins(1,4,5)P₃ produced with different concentrations of PE (e.g. Dora et al., 2008). The increased 5-phosphatase levels and lack of Ins(1,4,5)P₃-R on the VSMC side of the MEJ would ensure that Ins(1,4,5)P₃ from ECs does not affect VSMCs. It is clear that this metabolic aspect of the spatial heterocellular Ca²⁺ signaling will require more investigation.

Protein placement at the MEJ is not a random occurrence (e.g. Dora et al., 2008; Isakson et al., 2007; Isakson and Duling, 2005; Sandow et al., 2006). The data presented herein demonstrate selective localization of and functionally relevant Ins(1,4,5)P₃-R1 on the EC side of the MEJ, which might provide an explanation for the unidirectional response seen with Ins(1,4,5)P₃ between ECs and VSMCs and provide a mechanism by which heterocellular Ca²⁺ communication between the two cell types occurs. Taken together, the functional data imply that the MEJ is capable of being a unique cellular signaling domain in the vasculature.

All C57/Bl6 mice (Taconic) or Cx40^–/– mice were males between 6 to 10 weeks of age and used according to the University of Virginia Animal Care and Use Committee guidelines. Mice were euthanized with an intraperitoneal injection of 60-90 mg/kg pentobarbital.

For each isolation (at least three), ten cremaster muscles from five male mice were rapidly removed and pooled in ice-cold recovery buffer (HBSS, 5% BSA, 1% sodium pyruvate, 25 mM HEPES, 1% glutamine, 0.5 mM L-ascorbic acid and 0.1% fungizone). The cremasters were moved to isolation buffer [85% Ca²⁺, magnesium-free-HBSS, 15% DMEM, 15 μg/ml elastase (Sigma), 1.25 mg/ml collagenase VII (Sigma), 5% BSA, 1% sodium pyruvate, 25 mM HEPES, 1% glutamine, 0.5 mM L-ascorbic acid and 0.1% fungizone at 37°C] and vortexed before being put on a rotator in a 37°C oven for 8 minutes. After 8 minutes, the cremaster samples were removed and vortexed again for 30 seconds, then placed back in the oven. This process was repeated a total of four times. After the last vortex, the samples were rapidly strained over an 80-mesh sieve into an autoclaved beaker, rinsed with FBS and centrifuged at 25 g for 5 minutes. Upon supernatant removal, recovery buffer at 37°C was added back with 8×10⁷ CD31-coated magnetic beads (Dynal, Invitrogen) or NG2 (Murfee et al., 2005)-coated magnetic beads [made by incubating 1.34×10⁸ anti-rabbit magnetic beads (Dynal) with 100 μl rabbit anti-NG2 (1 μg/μl; Chemicon)]. The cells and the magnetic beads were rotated inside a 37°C incubator for 30 minutes and separated by a magnet. The supernatant was removed and culture medium (see below) was applied. This process yields approximately 1.1×10⁶ cells with the CD31-coated beads and 1.4×10⁶ cells with the NG2-coated beads. Trypan-blue exclusion assays were always greater than 90% or the isolated cells were discarded. These results demonstrated a highly enriched population of rapidly isolated cremasteric microvascular ECs and VSMCs (supplementary material Figs S1-S3; Fig. 1).

ECs were grown in MCDB-131 (Invitrogen) supplemented with 20% FBS, 1% glutamine (Invitrogen), 1% penicillin/streptomycin (Invitrogen), 1% sodium pyruvate (Invitrogen), 40 μg/ml heparin and 20 μg/ml endothelial cell growth supplement; VSMCs were grown in MEM supplemented with 10% FBS, 1% penicillin/streptomycin, 1% glutamine and 1% non-essential amino acids.

The cremasteric ECs and VSMCs were assembled into a VCCC as originally described (Isakson and Duling, 2005). VSMCs were plated at 1×10⁵ on the bottom of each Transwell insert (polyester, 0.4-μm pore diameter; Corning) before ECs were plated on the top of insert (1.5×10⁵).

Either the ECs or the VSMCs were loaded with the acetomethoxyester form of Fluo-4 (2.5 mM with 0.0025% pluronic F127 and 0.1% DMSO; all supplied by Invitrogen) and the VCCC was mounted on an Olympus FV200 confocal microscope (for details, see Isakson et al., 2007). The pixel intensity from each image was subtracted from a background image of the Transwell insert devoid of cells after experiments were finished (F_sub). Maximum fluorescence intensity (F_max) of both cell types was determined at the termination of each experiment by application of 10 mM ionomycin and stimulation with 100 mM ATP (Dora and Duling, 1998). Relative [Ca²⁺]_i values are plotted as the % F_max (Isakson et al., 2007). In all experiments, when stimulated ECs were used, 35 mM ATP was applied (Beny, 2004; Isakson et al., 2007) and, when stimulated VSMCs were used, 10 mM PE was applied (Isakson et al., 2007; Langlands and Diamond, 1990). Statistics for determining significance between EC and VSMC Ca²⁺ responses was at P<0.05 and determined by one-way ANOVA (Tukey post-hoc test); error bars are ± s.e.

Application of 18 α-GA (35 mM; Sigma) was performed as previously described (Isakson et al., 2007). The acetomethoxyester form of BAPTA [20 mM; Invitrogen (Yashiro and Duling, 2000)] was loaded into cells following loading of Fluo-4. Controls for BAPTA loading of a specific cell type were obtained by stimulating the unloaded cell type (e.g. Isakson et al., 2007) (also data not shown). The cell-permeate selective Ins(1,4,5)P₃-R blocker XPC [20 μM; Sigma (Oka et al., 2002)] was added to Fluo-4-loaded cells. Cells were loaded with 5-PI (K_i=4 μM; 500 μM dissolved in 0.5% DMSO; NMR indicated purity >95%, catalog number 524620, lot number 746693, Calbiochem) using a pinocytotic kit (Invitrogen) (e.g. Isakson and Duling, 2005) after loading with Fluo-4.

All siRNA was applied directly to either the ECs or VSMCs on the Transwell insert. Negative control siRNA (25 nM) and Cx43 siRNA duplexes (15 nM for both duplex sequences) have previously been verified (Isakson and Duling, 2005) and demonstrated to have no effect on other connexins. We used three different Ins(1,4,5)P₃-R duplex sequences per isoform (Table 1), transfected simultaneously (siLentFect, Bio-Rad) at 10 nM each. In each case, cells were used 48 hours after transfection.

Griffonia simplicifolia conjugated to Alexa Fluor 594, Helix pomatia conjugated to Alexa Fluor 594, acetylated low-density lipoprotein (LDL) conjugated to Alexa Fluor 594, phalloidin conjugated to Alexa Fluor 488 and all secondary antibodies were Alexa fluorphores obtained from Invitrogen. Desmin, Cx43, α-actin and β-actin (all monoclonal) were from Sigma; VE-cadherin and type-I 5-phosphatase were from Santa Cruz Biotechnologies; Cx40 and Cx37 were from ADI; Cx45 was a kind gift from Thomas H. Steinberg (Lecanda et al., 1998). Two anti-Ins(1,4,5)P₃-R1 antibodies were used [NeuroMab facility (UC DavisNINDS/NIMH), and Alomone] and anti-Ins(1,4,5)P₃-R2 and anti-Ins(1,4,5)P₃-R3 were obtained from Millipore. Owing to the reported cross-reactivity of anti-Ins(1,4,5)P₃-R antibodies, the specificity of these antibodies was tested by siRNA knock down (Figs 5 and 6). Anti-rabbit and anti-mouse 12-nm gold beads were from Jackson Laboratories.

Immunohistochemistry was performed as previously described on the VCCC (Isakson and Duling, 2005) and cremasteric arterioles (Isakson et al., 2008). Quantification of protein on actin bridges in mouse cremaster arterioles was performed as previously described (Isakson et al., 2008).

Mouse cremaster tissue was fixed with 4% paraformaldehyde and embedded in LR White. Ultra-thin sections were washed with Tris-buffer, blocked with 1% ovalbumin and viewed on a Zeiss 900 electron microscope.

Immunoblots [lysate buffer: PBS with 1% SDS, 1% protease inhibitor cocktail (Sigma), 50 mM NaF and 2 mM PMSF] were performed as previously described (Isakson and Duling, 2005) and quantified as described (Olsen et al., 2005) using ImageJ software (NIH) using at least three different preparations.

We thank Brian Duling for critical feedback; Scott R. Johnstone, Angela Best, Katherine Heberlein and Michael Rizzo for evaluations of the manuscript; and Susan Ramos for electron microscopy images. This work is supported by NIH ROI HL088554 (B.E.I.) and an American Heart Association Scientist Development Grant (B.E.I.).