There are five subtypes of muscarinic acetylcholine receptors (M₁ – M₅ mAChR) with distinct yet overlapping tissue distributions. Muscarinic receptors regulate a variety of physiological responses ranging from cardiac homeostasis to cholinergic signaling in the brain 1. A common feature of mAChRs, and in fact all GPCRs, is that both their expression and activation are tightly regulated. Agonist-promoted trafficking of mAChRs, and most other GPCRs can be broken down into five distinct phases: agonist-binding promotes G protein dissociation (I) from the receptor which allows phosphorylation of specific serine and threonine residues (II) on internal loops of the receptor by G protein receptor kinases (GRKs). This phosphorylation allows the binding of β-arrestins (III) which promotes homologous desensitization and subsequent internalization of the receptor into clathrin coated pits. Following internalization, the receptor can either be dephosphorylated and recycled (IV) to the cell surface or targeted for degradation (V) in proteasomes or lysosomes 1.

β-arrestins have emerged as a central control point in the trafficking of nearly all GPCRs 2. In addition to mediating desensitization of GPCRs, β-arrestin also participates in clathrin-dependent endocytosis of activated receptors by directly interacting with clathrin and the clathrin-associated adaptor AP-2 34. Once internalized, β-arrestins are involved in regulation of post-endocytotic trafficking and recently have been shown to function as scaffolding proteins interacting with cellular trafficking machinery 25.

Initial reports examining the role of β-arrestin, clathrin and dynamin indicated that agonist-promoted internalization of the M₂ mAChR proceeded via both arrestin-dependent and -independent pathways 67. Later work, however, indicated the essential role of β-arrestin in M₂ mAChR internalization, and suggested that all M₂ mAChR internalization, like other mAChR subtypes, is dynamin-dependent 89.

Two classes of GPCRs have been identified with respect to kinetics of receptor recycling and interaction with β-arrestins. Class A receptors, such as the β₂ adrenergic (β₂AR), dopamine D1A and endothelin 1A receptors recycle rapidly and dissociate from β-arrestin prior to receptor internalization. Class B receptors, such as the vasopressin 2 (V2R), angiotensin 1a and neurotensin 1 receptors, recycle slowly and internalize in a stable association with β-arrestin 1011. Recently, it has become clear that the classification of receptors as A or B directly correlates with patterns of β-arrestin ubiquitination and deubiquitination 12. Stimulation of Class A β₂ARs leads to transient ubiquitination of β-arrestin with deubiquitination occurring shortly (minutes) after internalization 12. In contrast, stimulation of Class B V2Rs leads to a stable ubiquitination of β-arrestin 12.

Ubiquitination of proteins is a signal for degradation that leads to delivery to and degradation of proteins in the 26S proteasome 13. There are a number of examples where ubiquitination has been shown to be involved in the regulation of GPCRs, including the opiod receptors 14, yeast pheromone receptor 15, human immunodeficiency virus co-receptor CXCR4 16, and β₂ ARs 17.

Class A receptors, which do not internalize with β-arrestin, display a pattern of transient β-arrestin ubiquitination. Class B receptors, on the other hand, do internalize with β-arrestin, and display a sustained β-arrestin ubiquitination pattern 12. Shenoy et al. 17 showed that agonist stimulation led to the ubiquitination of both β-arrestin and the β₂ AR. Their data suggested that it was the ubiquitination of β-arrestin that was involved in the initial receptor internalization step whereas ubiquitination of the receptor itself was required for subsequent receptor degradation.

In the present study, we wanted to address several questions regarding the role of β-arrestin ubiquitination in the agonist-promoted down-regulation of mAChRs. First we wanted to determine if there was a role for β-arrestin ubiquitination in receptor down-regulation. Collectively, our data support a role for β-arrestin in differential targeting of mAChRs towards down-regulation. In addition, the data indicate that this differential targeting may be controlled by preferential interaction of mAChR subtypes with the different β-arrestin subtypes which in turn may be mediated by specific patterns of β-arrestin ubiquitination.

For a number of years there has been some controversy as to the role of β-arrestin in M₂ mAChR specific internalization and down-regulation. Our recent publication 9 established the essential role of β-arrestin in the internalization of the M₂ mAChR. The current study is a follow up which not only addresses the role of β-arrestin in agonist-promoted mAChR down-regulation, but also indicates a differential role for β-arrestin in M₁ versus M₂ mAChR down-regulation. As in the previous study, we performed our down-regulation experiments in mouse embryonic fibroblasts (MEFs) derived from β-arrestin 1/2 double knockout mice (MEF KO1/2) 18. Experiments performed using this cell line offer a significant advantage over previous studies of the role of β-arrestin in GPCR internalization and down-regulation that used dominant-negative or knockdown strategies because they avoid any possible complications that could arise from the presence of low levels of endogenous arrestin proteins. These cells have been characterized to confirm the absence of mAChR expression using both PCR and radioligand binding assays 9.

To determine whether or not MEF cells could be used in the current study, we performed agonist-promoted down-regulation time courses for both the M₁ and M₂ mAChR subtypes. Total receptor numbers were measured using the non-selective membrane permeable muscarinic antagonist quinuclidinyl benzilate (QNB), which binds to both surface and intracellular pools of mAChR. Wild-type MEF cells that were transfected with either eGFP-M₁ mAChR or HA-M₂ mAChR resulted in similar expression levels (~4000 fmol/mg). After 24 hr, cells were treated with 1 mM carbachol for the indicated time. Maximal down-regulation of the M₂ mAChR occurred after 6 hr of stimulation (Figure 1). In contrast, maximal down-regulation of the M₁ subtype did not occur until after 12 hr of carbachol stimulation. In addition, M₂ mAChRs were maximally decreased by only 22% while the M₁ subtype was decreased by 55%. These results demonstrate that exogenously expressed M₁ and M₂ mAChRs undergo agonist-promoted down-regulation in MEFwt cells. The different time course and extent of down-regulation suggest the possibility that down-regulation of the two subtypes may involve distinct pathways. Despite the fact that M₂ mAChRs were maximally down-regulated by 6 hr of stimulation, subsequent single time point experiments used the 12 hr time point so that the M₁ and M₂ subtype experiments would be compared at a standardized time point.

Several recent studies have reported that mAChR down-regulation occurs independently of agonist-promoted internalization 192021. Recently, we demonstrated that agonist-promoted internalization of M₂ mAChRs in MEF cells was β-arrestin dependent 9. To examine whether or not agonist-promoted down-regulation of M₁ and M₂ mAChRs is β-arrestin dependent, we performed down-regulation studies in MEF β-arrestin double knockout cells transiently expressing either M₁ or M₂ mAChRs. MEF KO1/2 cells were transfected with eGFP-M₁ mAChR or HA-M₂ mAChR and either FLAG-β-arrestin 1 or 2. After 24 hr, cells were treated with 1 mM carbachol for 12 hr. In the absence of exogenous β-arrestin, there was no down-regulation in response to agonist stimulation in the double knockout cell line (Figure 2A–D). Expression of either β-arrestin 1 or β-arrestin 2, however, rescued down-regulation of both receptor subtypes in response to agonist (Figure 2A – D). These results demonstrate that agonist-promoted down-regulation of M₁ and M₂ mAChRs is β-arrestin dependent.

Recent studies have demonstrated that agonist stimulation of GPCRs results in ubiquitination of β-arrestin 2 12. This group demonstrated receptor specific agonist-promoted ubiquitination of β-arrestin 2 that was either transient (peaked at 1 min of agonist stimulation) or stable (still present after 15 min of agonist stimulation) 12. We performed experiments in order to determine if agonist stimulation of M₁ or M₂ mAChRs promoted the ubiquitination of β-arrestin and if so, examine whether the observed ubiquitination was transient or stable. MEF KO1/2 cells were transfected with FLAG-β-arrestin 2 and eGFP-M₁ or HA-M₂ mAChR and treated with carbachol for 0, 1, 3, 15 and 30 min. Stimulation of M₁ and M₂ mAChRs significantly increased ubiquitination of β-arrestin 2 (Figure 3). This increase is clearly visible at both the 15 and 30 min time point for both mAChR subtypes. These results demonstrate that stimulation of M₁ or M₂ mAChRs promotes a stable ubiquitination of β-arrestin 2.

Having established that β-arrestin is required for mAChR down-regulation and that muscarinic stimulation leads to the ubiquitination of β-arrestin, we wanted to examine if there was a direct role for β-arrestin ubiquitination in receptor down-regulation. Since ubiquitination is known to serve as a signal for protein degradation 13 we were interested in determining whether disruption of the ubiquitin/proteasome pathway would affect the agonist-promoted down-regulation of the mAChRs. To confirm that β-arrestin 2 promotes down-regulation via a ubiquitin dependent pathway, we examined agonist-promoted down-regulation of the M₂ mAChR in the presence of the proteosomal inhibitor lactacystin (impairs ubiquitin recycling). MEFwt cells were transfected with HA-M₂ mAChR and after 24 hr cells were incubated for 20 min with or without 10 μM lactacystin prior to treatment with 1 mM carbachol for 4 hr. Lactacystin was able to completely block agonist-promoted down-regulation (Figure 4) suggesting that the ubiquitination machinery has a role in mediating agonist-promoted down-regulation of mAChRs.

To determine whether or not the effects of lactacystin were occurring at the level of receptor internalization, we examined the effect of lactacystin on the agonist-promoted internalization of mAChRs using the membrane impermeable muscarinic antagonist N-methylscopolamine (NMS). MEFwt cells were transfected with HA-M₂ mAChR, and after 24 hr cells were incubated for 20 min in the absence or presence of 10 μM lactacystin prior to treatment with 1 mM carbachol for 30 min. Pretreatment with inhibitor had no effect on agonist-promoted internalization (Figure 5). These results indicate that agonist-promoted down-regulation of mAChRs involves receptor β-arrestin ubiquitination.

To examine the consequences of β-arrestin ubiquitination on receptor down-regulation, we expressed a yellow fluorescent protein-tagged β-arrestin 2-ubiquitin chimera that cannot be deubiquitinated by cellular deubiquitinases (YFP-β-arrestin 2-Ub) 12. MEF KO1/2 cells were transfected with eGFP-M₁ or HA-M₂ mAChR and either FLAG-β-arrestin 2 or YFP-β-arrestin 2-Ub. After 24 hr, cells were treated with 1 mM carbachol for 12 hr. There was a 40% agonist-promoted down-regulation of the M₁ mAChR by β-arrestin 2 alone which increased to 70% in the presence of β-arrestin 2-Ub (Figure 6). These values were 62 and 95%, respectively, for the M₂ subtype (Figure7). β-arrestin 2-Ub also increased the constitutive receptor down-regulation, and again the effect was much larger for the M₂ (90%) vs M₁ (50%) subtype (Figure 6A and 6B). It is clear from these data that ubiquitination enhanced the ability of β-arrestin 2 to mediate both constitutive and agonist-promoted down-regulation of both the M₁ and M₂ mAChRs.

Having demonstrated a role for ubiquitin in agonist-promoted mAChR degradation, we were interested in the effects of disrupting β-arrestin 2 ubiquitination on receptor down-regulation. Several lysine residues on β-arrestin are known to be sites of ubiquitination 22. To further confirm the essential role of ubiquitination in agonist-promoted down-regulation, we examined the ability of specific β-arrestin 2 lysine mutants to mediate agonist-promoted down-regulation of the M₁ and M₂ mAChR. MEF KO1/2 cells were transfected with eGFP-M₁ or HA-M₂ mAChR and either empty vector (control), FLAG-β-arrestin 2 (wild-type), FLAG-β-arrestin 2^{K18R, K107R, K108R, K207R, K296R} or FLAG-β-arrestin 2^{K11R, K12R}. After 24 hr, cells were treated with 1 mM carbachol for 12 hr. As shown previously, there was no down-regulation in the control cells in the absence of β-arrestin 2 (Figure 7). All three β-arrestin 2 constructs were able to rescue agonist-promoted down-regulation of the M₁ subtype. We also noted a large constitutive effect on down-regulation with both mutant β-arrestins. In contrast, for the M₂ subtype only wild-type β-arrestin 2 (24%) and β-arrestin 2^{K11R, K12R} (27%) were able to rescue agonist-promoted down-regulation (Figure 7). β-arrestin 2^{K18R, K107R, K108R, K207R, K296R}, had lost the ability to rescue agonist-promoted down-regulation of the M₂ mAChR (Figure 7).

Since it was possible that the effect of β-arrestin 2^{K18R, K107R, K108R, K207R, K296R} on M₂ mAChR down-regulation seen in the previous experiment was actually occurring at the receptor internalization step, we performed receptor internalization experiments with [³H]-NMS. Transfections were identical to the previous experiment. After 24 hr, cells were treated with 1 mM carbachol for 1 hr. All β-arrestin 2 constructs (wild-type or lysine mutants) were able to rescue agonist-promoted internalization of M₂ mAChRs in MEF KO1/2 (Figure 8).

Since β-arrestin 2^{K18R, K107R, K108R, K207R, K296R} not only appears to interfere with the agonist promoted down-regulation of the M₂ mAChR but also increased the constitutive down-regulation of the M₁ mAChR we examined the ability of this mutant to co-localize with the M₁/M₂ mAChR compared to wt and β-arrestin 2^{K11R, K12R} following agonist stimulation. MEF KO1/2 were plated on cover slips in 6-well plates. 24 hrs after transfection with eGFP-M₁ mAChR or HA-M₂ mAChR and FLAG tagged β-arrestin (wt or lysine mutants) cells were treated for 30 min or 12 hrs with 1 mM carbachol. Cells were fixed and processed for indirect immunofluorescence as described in methods. After 30 minutes of agonist exposure there is clear overlap of the M₂ mAChR receptor and β-arrestin signal signified by the yellow puncta (Figure 9) for the wt and β-arrestin 2^{K11, K12R}. This overlap is absent with β-arrestin 2^{K18R, K107R, K108R, K207R, K296R} which indicates a disrupted or impaired interaction between this β-arrestin mutant and the M₂ mAChR. Surprisingly, despite our previous observation of the effects of β-arrestin 2^{K18R, K107R, K108R, K207R, K296R} on constitutive M₁ down-regulation, no co-localization was observed between β-arrestin and the M₁ mAChR (Figure 10).

Finally we used immunocytochemistry with the lysosomal marker LAMP-1 to examine the effects of the β-arrestin 2^{K18R, K107R, K108R, K207R, K296R} mutant on mAChR lysosomal targeting down-regulation. MEF KO1/2 were transfected with HA-M₂ mAChR and either empty vector (control), FLAG-β-arrestin 2 (wild-type), FLAG-β-arrestin 2^{K11R, K12R} or FLAGβ-arrestin 2^{K18R, K107R, K108R, K207R, K296R}. After 24 hr, cells were treated with 1 mM carbachol for 6 hr. There was significant co-localization of the receptor with LAMP-1 in the presence of wild-type β-arrestin 2 or β-arrestin 2^{K11R, K12R} (Figure 11). It is clear, however, that β-arrestin 2^{K18R, K107R, K108R, K207R, K296R} shows reduced co-localization of the M₂ mAChR with LAMP (Figure 11).

Collectively, these data indicate that the agonist-promoted down-regulation of the M₁ and M₂ mAChRs involves differential sorting of the receptor by ubiquitinated β-arrestin.

Our recent work established the essential role of β-arrestin in the internalization of the M₂ mAChR 9. The present study extends the observations of previous work and demonstrates that the agonist-promoted down-regulation of M₁ and M₂ mAChRs is β-arrestin dependent, and that the ubiquitination pattern of β-arrestin has a critical role in the differential down-regulation for M₁ vs M₂ mAChRs.

It has been previously established in a variety of cell lines that a prolonged activation of M₁ or M₂ mAChRs induces receptor down-regulation. In agreement with these findings, long-term stimulation of M₁ or M₂ mAChRs induced receptor down-regulation with the M₁ mAChR showing significantly more down-regulation than M₂ subtype. This observation suggests that the two subtypes are differentially regulated by endogenous β-arrestins. No down-regulation of either mAChR subtype occurred in the absence of β-arrestin and either β-arrestin subtype was able to rescue receptor down-regulation.

The observations of Shenoy and co-workers suggest that the ubiquitination of β-arrestin is a critical control point in the trafficking of different classes of GPCRs 121722. The time course of ubiquitination of β-arrestin 2 in response to muscarinic stimulation in MEF KO1/2 cells transiently expressing the M₁ or M₂ mAChRs was slow in onset and stable over time. This observation suggests that both M₁ and M₂ mAChRs display a Class B β-arrestin ubiquitination pattern despite the fact that the M₁ mAChR is known to recycle rapidly compared to M₂ 23, and does not show a stable endocytotic co-localization with β-arrestin 2 9 – both characteristics of Class A GPCRs. This observation suggests that there is some flexibility in the Class A vs B distinction. This same flexibility has been observed in the categorization of somatostatin receptor subtypes 24.

To confirm that agonist-promoted down-regulation of mAChRs was ubiquitin-dependent we performed internalization/down-regulation experiments in the presence of the proteosomal inhibitor lactacystin. Lactacystin is a specific inhibitor of the 26S proteasome that functions by covalently modifying the active site and inhibiting the enzymatic activity of the proteasome 13. This inhibition of the enzyme would inhibit recycling and thus deplete the available ubiquitin. Lactacystin was able to completely block agonist-promoted down-regulation with no effect on agonist-promoted internalization. These results suggested that the availability of free ubiquitin has a role in agonist-promoted down-regulation of the mAChRs. This effect is most likely due to effects on ubiquitin-dependent lysosomal sorting as evidenced by a later experiment that demonstrated the co-localization of the agonist-internalized M₂ mAChR with the lysosomal membrane protein LAMP-1.

Having established that β-arrestin is ubiquitinated via mAChR activation, we then examined whether the ubiquitination of β-arrestin enhances its ability to mediate agonist-promoted down-regulation. M₁ mAChR levels were reduced and M₂ levels were nearly completely ablated in the presence of β-arrestin 2-Ub compared to wild-type β-arrestin. The effect was seen on both constitutive and agonist promoted down-regulation.

In order to examine the role of β-arrestin ubiquitination in targeting mAChRs toward degradation, we used β-arrestin lysine mutants that have been previously shown to have a role in the endocytotic trafficking of other GPCRs. The constructs, generated by Shenoy and co-workers, contain mutations of potential ubiquitination sites near the amino terminus (β-arrestin 2^{K11R, K12R}) and at five other sites in the protein (β-arrestin 2^{K18R, K107R, K108R, K207R, K296R}) 22. We were particularly interested in the possible differential effects of β-arrestin ubiquitination patterns on M₁ vs M₂ down-regulation given the fact that M₁ and M₂ mAChRs did not fit neatly into the classical Class A vs B distinction.

Both of the β-arrestin lysine mutants were able to mediate agonist-promoted down-regulation of the M₁ mAChR to the same extent as wild-type β-arrestin. While our data clearly indicates an essential role for β-arrestin ubiquitination in M₁ mAChR down-regulation, other lysine residues than the ones examined must be the targets for the necessary ubiquitination required for targeting the M₁ mAChRs toward down-regulation. We did observe increased constitutive down-regulation of the M₁ mAChR with both mutant β-arrestin constructs. It is possible that the pattern of ubiquitination of these mutant β-arrestins altered the constitutive targeting of the M₁ mAChR toward lysosomal down-regulation. This observation is one of several in this study that suggests that the pattern of ubiquitination on β-arrestin is directly implicated in the downstream targeting of mAChRs. Previously it has been suggested that the only role for β-arrestin ubiquitination is to promote the internalization of GPCRs.

We obtained surprisingly different results with the effects of the β-arrestin lysine mutants on the agonist-promoted down-regulation of the M₂ mAChR subtype. Mutation of the two lysine residues in β-arrestin 2^{K11R, K12R} had no effect on either agonist-promoted internalization or down-regulation of M₂ mAChRs. This result is similar to those seen with other Class B receptors such as the V2R and NK1R 22. The results are different however from those seen with another Class B receptor, the AT1a receptor. When this receptor was co-expressed with β-arrestin 2^{K11R, K12R}, agonist stimulation resulted in transient association of the receptor with β-arrestin 2, thus converting the receptor to a Class A type.

In contrast, the β-arrestin 2^{K18R, K107R, K108R, K207R, K296R} mutant significantly reduced M₂ mAChR degradation with no effect on receptor internalization. These results are similar to those seen with the Class B V2R, where β-arrestin 2^{K18R, K107R, K108R, K207R, K296R} was shown to impair association of V2Rs with β-arrestin 2 on endosomes 22.

The fact that β-arrestin 2^{K18R, K107R, K108R, K207R, K296R} was able to internalize but not down-regulate the M₂ mAChR suggests that one or more of these lysines has a role in down-regulation but not internalization for this subtype. These data suggest a direct role for the ubiquitination state of β-arrestin in the lysosomal targeting of the M₂ mAChR.

Unlike effects observed with the M₂ mAChR, β-arrestin 2^{K18R, K107R, K108R, K207R, K296R} was able to mediate down-regulation of M₁ mAChRs. This observation clearly demonstrates that the role of these lysines in down-regulation is subtype specific and that a specific pattern of ubiquitination on β-arrestin can differentially target the two mAChR subtypes for down-regulation.

Co-localization of wild-type and β-arrestin 2^{K11R, K12R} with the M₂ mAChR, observed after 30 min of agonist treatment is absent with β-arrestin 2^{K18R, K107R, K108R, K207R, K296R} which may indicate a disrupted or impaired interaction between this β-arrestin mutant and the M₂ mAChR. β-arrestin 2^{K18R, K107R, K108R, K207R, K296R} was also shown to disrupt the co-localization of the M₂ mAChR and the lysosomal marker LAMP-1 confirming our observation that β-arrestin 2^{K18R, K107R, K108R, K207R, K296R} does not mediate the agonist-promoted degradation of the M₂ mAChR. Surprisingly, no co-localization of M₁ mAChR was observed with any of the three β-arrestin constructs. Time points from 1 min to 12 hrs of carbachol treatment were performed to attempt to image co-localization between the M₁ mAChR and the β-arrestins (data not shown). This observation seems to indicate that the role of β-arrestin in M₁ mAChR down-regulation may involve a transient interaction between β-arrestin and the receptor which fates the receptor toward degradation.

Our observations of M₂ mAChR trafficking are different from the results of work with the Class A β₂ AR 17 which suggests that ubiquitination of β-arrestin is required for receptor internalization and that ubiquitination of the receptor itself is required for receptor degradation. Our data indicate that, for the M₁ mAChR, the ubiquitination state of β-arrestin appears to have a role in the level of constitutive (agonist-independent) down-regulation and for the M₂ mAChR the ubiquitination state of β-arrestin has a direct role in the agonist-promoted down-regulation of the receptor. Of the seven lysine residues we examined, at least one (or more) of the group 18, 107, 108, 207 and 296 are required for this M₂ mAChR down-regulation. We also observed that M₂ mAChR internalization was independent of β-arrestin ubiquitination at these same lysine residues that have been shown to interfere with the association of β-arrestin with other Class B receptors 22. The possibility that other lysine residues than those we examined are ubiquitinated in order to internalize the receptor would preserve the role for β-arrestin in receptor internalization. The absence of down-regulation of M₂ mAChRs with β-arrestin 2^{K18R, K107R, K108R, K207R, K296R} however, clearly indicates a direct role for the ubiquitination of β-arrestin in receptor down-regulation.

We conclude that agonist-promoted down-regulation of both M₁ and M₂ mAChRs is β-arrestin-dependent. We further conclude that this down-regulation is specifically modulated by the ubiquitination state of β-arrestin and that this β-arrestin ubiquitination differentially targets the receptor subtypes for degradation in lysosomes. Significantly, where it has been suggested that Class A GPCR down-regulation proceeds primarily via receptor ubiquitination, these data indicate that the ubiquitination state of β-arrestin 2 has an essential role in the differential targeting the M₁ and M₂ mAChR toward down-regulation. Future studies will examine the role of receptor ubiquitination in M₁ and M₂ mAChR down-regulation. Specifically, we will examine whether expression of wild type vs the β-arrestin 2^{K18R, K107R, K108R, K207R, K296R} mutant differentially affects M₁ and M₂ mAChR ubiquitination and subsequent down-regulation.

AT1aR: angiotensin 1a receptor; β₂ AR: β₂ adrenergic receptor; CHO: Chinese hamster ovary cells; GPCR: G protein-coupled receptor; MAb: monoclonal antibody; mAChR: muscarinic acetylcholine receptor; MEF: mouse embryonic fibroblast; NK1R: neurokinin 1 receptor; NMS: N-methylscopolamine; PAb: polyclonal antibody; QNB: quinuclidinyl benzilate; V2R: vasopressin 2 receptor.

The authors declare that they have no competing interests.

VAM performed down-regulation, internalization, immunoblotting, participated in the design of the study and drafted the manuscript. KJTperformed down-regulation and immunocytochemistry. KMH performed immunocytochemistr. NAM was involved in critical revision of the manuscript. DAJ designed the study and participated in drafting and critical evaluation of the manuscript. All authors read and approved the final manuscript.