Content uploaded by Cynthia D Cooper
Author content
All content in this area was uploaded by Cynthia D Cooper on Jun 01, 2014
Content may be subject to copyright.
Analysis of Phosphorylation of Connexin43 at S325/328/330 in
Normoxic and Ischemic Heart
Paul D. Lampe
1,*
, Cynthia D. Cooper
1
, Timothy J. King
1,2
, and Janis M. Burt
3
1 Molecular Diagnostics Program, Fred Hutchinson Cancer Research Center and Department of
Pathobiology, University of Washington, Seattle, WA
3 Department of Physiology, University of Arizona, Tucson, AZ
Summary
The functional consequences of Connexin43 (Cx43) phosphorylation remain largely unexplored.
Using an antibody that specifically recognizes Cx43 phosphorylated at serines 325/328/330
(pS325/328/330), we show that this form of Cx43 as well as total Cx43 labeling was restricted to the
intercalated disk region of normal ventricular tissue. In ischemic heart, significant relocalization of
total Cx43 to the lateral edges of myocytes was evident; however pS325/328/330-Cx43 remained
predominately at the intercalated disk. Western immunoblots indicated a 8-fold decrease in
pS325/328/330-Cx43 in ischemic tissue. Peptide binding and competition experiments indicated that
our antibody mainly detected Cx43 phosphorylated at S328 and/or S330 in heart tissue. To evaluate
how this change in Cx43 phosphorylation might contribute to ischemia-induced downregulation of
intercellular communication, we stably transfected Cx43
−/−
cells with Cx43-S325/328/330A (Cx43-
TM). Cx43-TM was not efficiently processed to isoforms that have been correlated with gap junction
assembly. Nevertheless, Cx43-TM cells were electrically coupled, although coupling developed over
a slowed time course. Fully open channels were only rarely observed in the Cx43-TM cells, and
Lucifer Yellow dye coupling was significantly reduced compared to wild-type. These data suggest
that phosphorylation of Cx43 at serines 325, 328 and/or 330 may influence channel permselectivity
and regulate the efficiency of gap junction assembly.
Keywords
Connexin43; Gap junction; heart; ischemia; phosphorylation
Introduction
Gap junction mediated intercellular communication facilitates direct communication among
adjacent cells by allowing passage of ions and small metabolites (White and Paul, 1999; Saez
et al., 2003; Sohl and Willecke, 2004). Vertebrate gap junctions, composed of integral
membrane proteins from the Connexin gene family, are critically important in regulating
embryonic development, coordinated contraction of excitable cells, tissue homeostasis, normal
cell growth and differentiation (Saez et al., 2003; Sohl and Willecke, 2004). Furthermore,
connexin mutations have been linked to several diseases (Bergoffen et al., 1993; Gong et al.,
1997; Kelsell et al., 1997) including oculodentodigital dysplasia, a disease linked to
Connexin43 (Cx43) mutations that can cause atrioseptal defects and arrhythmias (Paznekas et
al., 2003). Twenty-one connexin genes have been identified in humans (Sohl and Willecke,
*to whom correspondence should be addressed: Paul D. Lampe, Ph.D., Fred Hutchinson Cancer Research Center, 1100 Fairview Avenue
N., M5C800, P.O. Box 19024, Seattle, WA 98109, plampe@fhcrc.org, P: 206-667-4123, F: 206-667-2537.
2
Current Address: Hawaii Biotech, Inc., Aiea, HI
NIH Public Access
Author Manuscript
J Cell Sci. Author manuscript; available in PMC 2007 February 15.
Published in final edited form as:
J Cell Sci. 2006 August 15; 119(Pt 16): 3435–3442.
NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author Manuscript
2004). During intercellular channel formation, six connexin proteins oligomerize into a hemi-
channel or connexon; connexons are then transported to the plasma membrane by as yet
unknown mechanisms. The intact channel is formed when one hemi-channel docks with a
second in an opposing cell. Once assembled, groups of these intercellular channels (termed
gap junctional plaques) mediate the diffusion of ions, amino acids, second messengers and
other metabolites between the cytoplasms of the two cells (White and Paul, 1999; Sohl and
Willecke, 2004). The channels can be gated in response to various stimuli, including changes
in voltage, pH and connexin phosphorylation. Regulation of gap junctional communication
could occur by controlling any one of the steps mentioned above, however, many of the
regulatory mechanisms underlying these events remain elusive.
Cx43, the most ubiquitously expressed connexin, is differentially phosphorylated at a dozen
or more serine residues throughout its life cycle (Lampe and Lau, 2004). Cx43 from cultured
cells commonly demonstrates multiple electrophoretic isoforms when analyzed by SDS-
PAGE: a faster migrating form (sometimes referred to as P0 or NP) that includes the non-
phosphorylated isoform, and multiple slower migrating forms (sometimes termed P1 and P2
(Musil and Goodenough, 1991)). Following alkaline phosphatase treatment, the
phosphorylated species collapse to the fastest migrating form, suggesting that phosphorylation
is the primary covalent modification detected in SDS-PAGE analysis although no assignment
of specific phosphorylation sites to a change in Cx43 migration has been made. Pulse-chase
studies using Brefeldin A indicate some Cx43 phosphorylation occurs prior to reaching the
plasma membrane (Laird et al., 1995). This phosphorylation event may be necessary for
maintaining hemichannels in their closed state until docking occurs (Bao et al., 2004). In
addition, studies investigating phosphorylation in normal rat kidney (NRK) cells show that
Cx43 acquires resistance to Triton X-100 once it has been phosphorylated to the slower
migrating isoforms and assembled into gap junction plaques (Musil and Goodenough, 1991).
Thus, uncharacterized phosphorylation events have been correlated with changes in assembly,
acquisition of Triton X-100 insolubility and, potentially, degradation of Cx43 gap junction
channels.
In the normally functioning ventricle, Cx43 is localized to intercalated disks where it supports
the longitudinal and transverse spread of the action potential resulting in coordinated contractile
activation. Myocardial ischemia leads to Cx43 “dephosphorylation” and loss of localization to
the intercalated disk, which likely contributes to contractile failure and arrhythmias (Beardslee
et al., 2000; Schulz et al., 2003). Casein kinase 1 (CK1) mediates phosphorylation of Cx43 at
S325/328/330 in vitro. In cultured cells these sites are routinely phosphorylated; inhibition of
CK1 reduces phosphorylation at these sites and reduces gap junction assembly (Cooper and
Lampe, 2002). We sought to determine here whether phosphorylation at these sites occurs in
heart tissue and whether this phosphorylation event is affected during ischemia. Using an
antibody specific for Cx43 when it is phosphorylated at S325/328/330, we show that Cx43
localized to intercalated disks is phosphorylated at one or more of these residues (likely S328
and S330) and that ischemia leads to loss of this phosphorylation and relocalization of the
protein. Furthermore, we show that mutation of these sites increases the migration of the Cx43
phospho-isoforms on SDS-PAGE, reduces the extent of gap junction formation and reduces
gap junctional communication.
Methods
Antibodies and Reagents
All general chemicals, unless otherwise noted, were purchased from Fisher Scientific. Mouse
anti-Cx43 antibodies, Cx43CT1, Cx43CT2, and Cx43IF1 were prepared against amino acids
360–382 of Cx43 and antibody Cx43NT1 against amino acids 1–20 (Cx43NT1) of Cx43 at
the Fred Hutchinson Cancer Research Center Hybridoma Development Facility (Seattle, WA).
Lampe et al. Page 2
J Cell Sci. Author manuscript; available in PMC 2007 February 15.
NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author Manuscript
A rabbit antibody against Cx43 was purchased from Sigma (St. Louis, MO, C6219). We made
a rabbit anti-pS325/328/330-Cx43 antibody by custom commercial preparation (ProSci Inc.,
Poway, CA; 13 week schedule) against a synthetic peptide that was phosphorylated at all 3
serines (CQAG(pS)TI(pS)N(pS)HAQ-amide) that had been linked via the N-terminal cysteine
to maleimide-activated KLH (Pierce Biotechnology, Rockford, IL) according to
manufacturer’s instructions. Phosphospecific antibody was affinity purified by linking to the
phosphorylated peptide and the corresponding nonphosphorylated version to SulfoLink
Coupling Gel (Pierce) and passing the serum first over a 3 ml column prepared with the
nonphosphorylated peptide gel followed by a phosphopeptide-SulfoLink column according to
the manufacturer’s instructions. After washing, phosphospecific antibody was eluted from the
column with 5% acetic acid and fractions were neutralized with ammonium acetate. Antibody
specific for phosphorylation at S325 and S328 was affinity purified by linking peptides
phosphorylated at S325/328 and S328/330 (CGQAG(pS)TI(pS)NSH-amide and CGSTI(pS)
N(pS)HAQ-amide) to SulfoLink beads to make 1.5 ml columns of each. The polyclonal
antibody was then run first over the pS328/330 column and the unbound antibody was run over
the pS325/328 column. The pS325/328 specific antibody that bound was then eluted as
described above. All antibodies were aliquoted and stored frozen at −80°C.
Cell culture
Cx43 knockout fibroblasts lacking Cx43 (clone 23-3) or exogenously expressing wild-type
Cx43 (Cx43-WT; clone 22C-3) were provided by Dr. Erica TenBroek (University of
Minnesota). Normal Rat Kidney cells (E51) were obtained from American Tissue Type
Collection (Manassas, VA). Cells were cultured in Dulbeccos Minimal Essential Medium
(DMEM, Fisher Scientific, Pittsburgh PA) supplemented with 5–10% fetal calf serum and
antibiotics (100 U/mL penicillin G and 100μg/mL streptomycin) in a humidified 5% CO
2
environment. Cx43 with serines at 325, 328 and 330 mutated to alanine (Cx43-TM), kindly
provided in pcDNA3.1 by Dr. Steven Taffet (SUNY Health Science Center, Syracuse NY),
was subcloned into pIREShygro. To make Cx43-TM expressing fibroblasts, pIREShygro
Cx43-TM was transfected into Cx43 knockout fibroblasts (clone 23-3) using lipofectAMINE
Plus transfection reagent (GibcoBRL/Invitrogen, Carlsbad CA) according to the
manufacturer’s instructions. Cells were selected at 500μg/ml Hygromycin B and were dilution
cloned in DMEM supplemented with hygromycin and endothelial cell growth supplement
(ECGS, 5mg/mL) to facilitate selection and growth of cells. Three separate clones that stably
expressed Cx43-TM were isolated (TM-A, TM-B, TM-C).
Immunodetection of Cx43 in Heart
Mouse studies were conducted under FHCRC Institutional Animal Care and Use Committee
approval. Inbred mice (11 months of age in a FVB/N:C57BL6 background) were anaesthetized
(avertin, 0.1mL/3g body weight) and hearts were excised and placed either in ice cold PBS
(calcium and glucose-free) for 30 seconds (control group) or incubated without coronary
perfusion in non-oxygenated, glucose-free PBS for 30 minutes at 37°C. After 30 seconds
(control) or 30 minutes (“ischemic”) of incubation, hearts were longitudinally bisected and
either sonicated in Laemmli sample buffer (for Western analysis) or fixed overnight at 4°C in
10% neutral-buffered formalin (for immunohistochemistry). Although not thoroughly
characterized, the 30-minute “ischemic” period in the presence or absence of 1 mM calcium
reproduces the effects of ischemia on Cx43 electrophoretic mobility and gap junction
remodeling (see results) demonstrated in better characterized models of ischemia.
Formalin-fixed tissue was processed, sectioned, immunostained and microscopically analyzed
as previously described (King and Lampe, 2004). Briefly, tissue sections were deparaffinized,
antigen retrieved, blocked and detected utilizing rabbit anti-Cx43 (1:250, Sigma) or rabbit anti-
pS325/328/330 (1:1000). Slides were washed and incubated with a biotinylated anti-rabbit
Lampe et al. Page 3
J Cell Sci. Author manuscript; available in PMC 2007 February 15.
NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author Manuscript
secondary antibody (1:250) and detected with ABC-avidin/biotin conjugate (Vectastain,
Vector Labs, Burlingame, CA).
Immunoblotting and Immunofluorescence
Approximately the same amount of cardiac tissue for each treatment was sonicated in sample
buffer supplemented with 50mM NaF, 1mM Na
3
VO
4
, 5% β-mercaptoethanol, 1 mM PMSF
and 1x Complete protease inhibitors (Roche Molecular Biochemicals, Alameda, CA).
Insoluble material was removed by centrifugation, and the soluble fraction was separated on
SDS-10% PAGE. After immunoblotting, protein was detected with a rabbit antibody against
pS325/328/330-Cx43 and a mouse anti-Cx43 (Cx43NT). The blots were also probed with anti-
GAPDH (Ambion, Austin, TX) and anti-Vinculin (Sigma) to confirm consistent loading.
Primary antibodies were visualized with fluorescent dye-labeled secondary antibodies
[AlexaFluor 680 goat anti-rabbit (Molecular Probes, Eugene, OR) and IRDye800-conjugated
donkey anti-mouse IgG (Rockland Immunochemicals)] and directly quantified using the LI-
COR Biosciences Odyssey infrared imaging system (Lincoln, NE) and associated software
(inverted images are shown).
To test the phosphospecificity of the p325/328/330 antibody, we treated with alkaline
phosphatase or left untreated equal amounts of protein lysates from fibroblasts expressing
Cx43-WT and Cx43-TM (clone A) as previously reported (Lampe et al., 1998), ran the lysates
in SDS-PAGE and performed a Western blot with Cx43NT1 and the pS325/328/330 antibodies
on the same blot as described above. To determine the specificity of binding of the antibody,
peptides representing singly phosphorylated S325, 328 and 330 (CGQAG(pS)TISN-amide,
CAGSTI(pS)NSHA-amide, CGSTISN(pS)HAQP-amide, respectively), doubly
phosphorylated S325/328 and S328/330 (CGQAG(pS)TI(pS)NSH-amide and CGSTI(pS)N
(pS)HAQ-amide) and the triply phosphorylated immunizing peptide were synthesized and
individually incubated (at 200 pg/ml) with pS325/328/330 antibody for 30 minutes prior to
probing Western blots of whole cell lysates from mouse heart (prepared and immunodetected
as described above) using a Surf-Blot apparatus (Idea Scientific, Minneapolis, MN) for the
primary antibody incubation step. The level of antibody binding was determined as described
above by LI-COR densitometry and normalized to the signal in the absence of peptide. The
amount of antibody that can bind to the different peptides was determined by linking the 6
peptides to Reacti-Bind Maleimide-activated 96-well plates (Pierce) at saturating
concentrations. The plates were washed, blocked and incubated with pS325/328/330 antibody
or the further purified pS325/328 specific antibody in quadruplicate according to
manufacturer’s instructions and with peroxidase labeled donkey anti-rabbit secondary antibody
(Chemicon, Temecula, CA). Peroxidase levels were detected using ABTS Peroxidase substrate
(KPL, Gaithersburg, MD). The amount of antibody bound was determined by normalizing to
the total Abs410nm for the immunizing peptide and mean and standard deviation were
calculated.
To test the effect of mutation at S325/328/330, fibroblasts exogenously expressing wild-type
Cx43 or Cx43-TM were grown to 90–95% confluency and cells were harvested in 1% Triton
X-100 in PBS supplemented with 50mM NaF, 1mM Na
3
VO
4
and protease inhibitors. These
samples were separated into Triton soluble and insoluble fractions by centrifugation at
13,000g at 4°C for 10 minutes. Triton insoluble fractions (pellets) were resuspended in
Laemmli sample buffer and sonicated. Duplicate parallel cultures were lysed in 1X Laemmli
SDS-sample buffer supplemented with phosphatase/protease inhibitors and 5% β-
mercaptoethanol (whole cell lysate), followed by brief sonication. Protein assays for equal
loading, electrophoresis and immunodetection using Cx43CT2 antibody was performed as
previously described (Cooper and Lampe, 2002).
Lampe et al. Page 4
J Cell Sci. Author manuscript; available in PMC 2007 February 15.
NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author Manuscript
Immunofluorescence was performed as previously described (Solan et al., 2003). Cx43 was
detected with anti-Cx43 antibody (C6219) and Alexa 594 anti-rabbit secondary antibody.
DAPI was added to visualize cellular nuclei. The coverslips were mounted onto slides with
DABCO antifade medium [25mg/ml of 1,4-Diazobicyclo-(2,2,2)octane (Sigma) diluted in
Spectroglycerol (Eastman Kodak Co.) and 10% PBS, pH 8.6] and viewed with a Nikon Diaphot
TE300 fluorescence microscope, equipped with a 40x (1.3 n.a.) oil objective and a Princeton
Instruments digital camera driven by an attached PC and Metamorph imaging software.
Dye Injection
Cx43-WT or Cx43-TM expressing cells were grown in 35mm dishes to 70–80% confluency.
Donor cells were microinjected with Lucifer yellow (1mg/ml dissolved in 0.15M LiCl) and
allowed to transfer dye for three minutes. Digital images were collected at identical camera
settings on the Nikon Diaphot TE300 fluorescence microscope described above. Following
imaging, the number of recipient cells for both conditions was quantified in a blind manner.
Junctional Conductance and Single Channel Activity
Confluent cells were trypsinized and replated at low density on glass coverslips. After 2–5h
coverslips were mounted in a custom made chamber and cells were visualized on an inverted
(Olympus IMT2) microscope equipped for phase contrast observation. Cells were bathed in
solution containing (in mM): 142.5 NaCl, 4 KCl, 1 MgCl
2
, 5 glucose, 2 sodium pyruvate, 10
HEPES, 1 BaCl
2
, 1 CaCl
2
, 15 CsCl, and 10 TEACl, pH 7.2 and osmolarity 315 mOsM.
Junctional conductance was determined using dual whole-cell voltage clamp techniques as
previously described (Cottrell et al., 2003). The pipette solution contained (in mM): 120 KCl,
14 CsCl, 3 MgCl
2
, 5 glucose, 9 HEPES, 9 EGTA, 9 TEACl, 5 Na
2
ATP, 30 KOH to adjust pH
to 7.2, 315 mOsM. Because Cx45 is sometimes detected at low levels in the parental Cx43−/
− cell line (Martyn et al, 1997), we first evaluated each cell pair for voltage dependent loss of
junctional conductance. Junctions displaying significant voltage dependent gating at
transjunctional voltages of 40mV (>10–15% decrease in conductance in 5–10s) were not
studied further. For voltage insensitive junctions, junctional conductance was evaluated with
10mV transjunctional pulses and single channel events were visualized with a transjunctional
driving force of 40mV following superfusion with 4mM halothane (Burt and Spray, 1989).
Event amplitudes were measured by hand for each cell pair, binned into 6.25pS bins, the binned
data from all pairs pooled, and the relative frequencies of each bin calculated and plotted. Data
were fit using Origin software as previously described (Cottrell et al., 2003).
RESULTS
Ischemia-induced “dephosphorylation” of Cx43
Previous studies have shown that the electrophoretic mobility of Cx43 isolated from ischemic
heart is increased, such that most of the protein migrates at a position that includes
dephosphorylated Cx43 so the assumption has been made that these changes in electrophoretic
mobility represent a reduction in phosphorylation at undefined sites (Beardslee et al., 2000;
Jain et al., 2003). In addition, inhibition of CK1 has been shown to reduce gap junction
assembly and to phosphorylate Cx43 in vitro at one or more of serines 325/328/330 (Cooper
and Lampe, 2002). We hypothesized that phosphorylation at sites necessary for efficient
assembly of channels might be compromised in the ischemic setting. To evaluate whether these
phosphorylation events occur in control and ischemic heart tissue, we generated a polyclonal
antibody that uniquely detects Cx43 phosphorylated at S325/328/330 (pS325/328/330–the
specificity of which is described in the next section).
In agreement with previous studies, the electrophoretic mobility of total Cx43 from ischemic
heart was increased compared to control possibly indicating unspecified “dephosphorylation”
Lampe et al. Page 5
J Cell Sci. Author manuscript; available in PMC 2007 February 15.
NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author Manuscript
event(s) had occurred (Fig. 1, each lane represents a distinct heart). When detected with the
phosphospecific antibody, we found that the pS325/328/330-Cx43 content decreased ~8 fold
(p<0.03) in ischemic hearts (6-fold when the loss of total Cx43 was taken into account).
The cellular distribution of Cx43 detected with the phosphospecific antibody and one to total
Cx43 varied in ischemic vs. normal tissue. Fig. 2 shows that in normal hearts virtually all total
Cx43 and pS325/328/330-Cx43 were localized to intercalated disks. As shown previously
(Severs et al., 2004), ischemia caused a redistribution of total Cx43 from the intercalated disk
to the lateral borders of myocytes. The pS325/328/330 signal was reduced at the intercalated
disk of ischemic heart tissue, consistent with the total Cx43 reduction, but, surprisingly,
pS325/328/330-Cx43 was undetectable at the lateral borders. If phosphorylation at these sites
promotes gap junctional functionality, then transverse conduction in the ischemic heart would
not be expected to increase (dramatically) despite increased Cx43 at lateral borders because of
the reduced phosphorylation at the S325/328/330 sites and consequent poor gap junction
functionality.
Analysis of the specificity of the pS325/328/330 antibody
To test the specificity of the pS325/328/330 antibody and to determine whether
phosphorylation at S325/328/330 directly affects protein localization and gap junction
formation, we transfected constructs encoding either wild-type Cx43 (WT) or Cx43 with serine
to alanine mutations at residues 325, 328 and 330 (Cx43-TM) into fibroblasts derived from
Cx43
−/−
mice and examined the properties of the gap junctions formed by these cells. We first
validated the phosphospecific nature of the antibody by examining its binding to extracts from
cells expressing Cx43-WT or Cx43-TM both before and after treatment with alkaline
phosphatase. As shown in Fig. 3A, the antibody to total Cx43 bound to 41–44 kDa bands in
all cases and revealed the classic increase in mobility to the fastest migrating position upon
alkaline phosphatase treatment. In contrast, pS325/328/330-specific antibody bound to wild-
type Cx43 only in the absence of alkaline phosphatase treatment and did not bind to the Cx43-
TM under any conditions. Furthermore, pS325/328/330 antibody appeared to bind only to the
slowest migrating isoform of Cx43; the Cx43-TM expressing cells also did not efficiently
convert Cx43 to the slower migrating phosphorylated isoforms.
To examine the specificity of the pS325/328/330 antibody and determine which sites are likely
phosphorylated in vivo, we synthesized 3 peptides with S325, 328 and 330 each singly
phosphorylated, two peptides, S325/328 and S328/330, doubly phosphorylated, and the
immunogen peptide triply phosphorylated and covalently bound these peptides to an activated
96-well plate. The p325/328/330 antibody was incubated in the well and a standard enzyme-
linked immunoabsorbent assay (ELISA) was performed (Fig. 3B, open bars). The pS328/330
peptide bound almost as much total antibody as the immunizing (triply phosphorylated) peptide
indicating that most of the antibody bound an epitope centered on these sites. The singly
phosphorylated peptides bound only a small amount of antibody. To further answer the question
of the specificity of this antibody preparation, we removed the antibody that bound to the
p328/330 peptide and purified that specific for pS325/328. Analysis of the binding of this
preparation to the 6 peptides in the ELISA format indicated that the pS325/328 purified
antibody indeed bound to mostly to the pS325/328 peptide followed by the triply
phosphorylated and the pS325 singly phosphorylated peptides (data not shown). These results
indicate that the antibody preparation is phosphospecific but polyclonal with reactivity to all
3 sites. To examine the ability of the 6 peptides to block antibody binding to Cx43, we
preincubated these peptides with the pS325/328/330 antibody prior to Western blot analysis
of heart lysates. Neither the pS325 nor pS328 singly phosphorylated peptides interfered
significantly with antibody binding to intact Cx43 leaving most of the signal remaining,
whereas the pS330 singly and all of the more extensively phosphorylated peptides did reduce
Lampe et al. Page 6
J Cell Sci. Author manuscript; available in PMC 2007 February 15.
NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author Manuscript
the Western signal to close to background levels (Fig. 3B, filled bars). Further, we tested the
ability of the pS325/328 specific antibody and found that it could bind to Cx43 prepared from
heart in an immunoblot but much less extensively than the antibody prior to the removal of the
pS328/330 reactivity (data not shown). Combined these antibody binding and competition
results indicate that this polyclonal antibody can bind to Cx43 phosphorylated at S325, S328
and S330 and that Cx43 is phosphorylated at S328 and/or S330 and to a lesser extent at S325
in heart tissue.
The role of phosphorylation at S325/328/330 in gap junction assembly
To be more quantitative about the extent of gap junction formation, we assayed the ability of
three separate Cx43-TM clones to phosphorylate Cx43 to the P2 isoform and incorporate it
into a Triton X-100 insoluble fraction, a hallmark of conventional gap junctions (Musil and
Goodenough, 1991). Like most cell lines, fibroblasts expressing wild-type Cx43 showed
several isoforms of Cx43 in the whole cell lysate and a dramatic enrichment of the P2 isoform
in the Triton insoluble fraction (Fig. 4, WT lanes). Transfectants expressing Cx43 with the TM
mutations (clones A, B, C) showed multiple differences in their Cx43 migration. In the whole
cell fraction, Cx43-TM migrated primarily at P0 with a second band between P0 and P1 and
no apparent P2. There was a significant reduction in the total amount of Triton insoluble Cx43
in clone B and C and essentially no bands that co-migrated with P2 were observed in any of
the clones. Variation in total Cx43 protein levels most likely represents clonal heterogeneity
among individual Cx43 expressing clones. Since the TM-A clone showed comparable levels
of total Cx43 expression to the Cx43-WT expressing cells, we used it for most of the subsequent
studies. We conclude that phosphorylation of Cx43 at S325, 328 and/or 330 is necessary for
Cx43 to migrate as the P2 isoform.
The localization of Cx43 in the Cx43-TM expressing clones was also examined via
immunofluorescence. In Fig. 5, we show that fibroblasts transfected with wild-type Cx43
showed both intracellular and punctate, gap junctional plaque-associated Cx43 (compare to
the lack of signal in the KO panel). In contrast, the TM mutant cell lines showed predominately
intracellular Cx43 although occasional apparent appositional labeling was observed (TM-A
clone is shown).
We next determined whether cells expressing Cx43-TM supported intercellular
communication. Parental Cx43 knockout fibroblasts (KO), wild-type Cx43 expressing
fibroblasts (WT) and the three individual Cx43-TM expressing clones A-C were microinjected
with Lucifer yellow dye and evaluated for dye transfer 3 minutes later. Digital images were
taken and the number of recipient cells was quantified. As shown in Fig. 6, the number of cells
receiving dye from donor cells was 60–85% lower in the Cx43-TM clones than for Cx43-WT
expressing cells and was comparable to the coupling observed in the parental fibroblasts that
lacked Cx43. Although Cx43 protein expression levels varied between the clones (see Fig. 4),
all 3 Cx43-TM clones including TM-A, the one that expressed similar levels of Cx43, showed
significant reductions in communication compared to the wild-type control (p<0.02). These
data suggest that phosphorylation at S325, S328 and/or S330 is critical for efficient dye transfer
in mouse fibroblasts.
We next evaluated whether junctional conductance and channel behavior differed as a
consequence of the serine to alanine mutations at the S325/328/330 sites. Based on the dye-
coupling data presented above, we expected junctional conductance to be significantly reduced
in the Cx43-TM cell clones. This was indeed the case for Cx43-TM cells plated for an
equivalent period of time as wild-type cells (2–5h) - junctional conductance in Cx43-TM cells
was 3.5±0.7 nS (n=24) vs. 6.2 ±1.1 nS (n=35, p<0.05) in Cx43-WT cells. Coupling between
Cx43-TM cells improved with time such that 24–30h after plating, conductance was 4.9±0.8
nS (n=28), a value not significantly different from Cx43-WT cells at 2–5h post plating. The
Lampe et al. Page 7
J Cell Sci. Author manuscript; available in PMC 2007 February 15.
NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author Manuscript
profile of channel conductances in the Cx43-TM cells was substantially different from that
observed in wild-type cells as shown in Fig. 7A and emphasized in the difference plot (Fig.
7B), with a severely decreased frequency for activity of fully-open channels (~110pS) and
increased frequency of 60–85pS channels.
Discussion
Previous work has established that myocardial ischemia rapidly induces uncharacterized
changes in Cx43 phosphorylation and localization, however no mechanistic connections
between specific phosphorylation events and changes in conduction or localization have been
identified. Here, we show that phosphorylation of Cx43 at serines 325, 328, and/or 330 was
drastically reduced in the ischemic heart. Further, we demonstrate that Cx43 localized to the
lateral borders of myocytes was not phosphorylated at these sites, whereas the Cx43 that
remained at the intercalated disk retained these modifications. Finally, we show that when these
sites were not phosphorylated, the event frequency for fully open channels, the overall
junctional conductance and the extent of Lucifer Yellow dye coupling were all significantly
reduced.
Although direct Cx43 phosphorylation on serines has long been correlated with gap junction
formation and function under basal conditions (Musil and Goodenough, 1991), few
mechanistic connections to specific phosphorylation events have been drawn. For instance,
phosphorylation at S364 was shown to be important for cAMP induced up-regulation of gap
junction assembly. However, cells expressing Cx43 with a S364A mutation assembled Cx43
into junctions nearly as well as wild-type expressing cells, suggesting S364 was not necessary
for assembly in homeostatic cells (TenBroek et al., 2001). Similarly, several serines are known
to be involved in mitogen-induced downregulation of communication including serines 255,
279 and 282 (MAPK) and serines 262 and 368 (protein kinase C), but these phosphorylation
events have been reported to negatively affect channel gating properties (Lampe et al., 2000;
Cottrell et al., 2003). S325, 328 and/or 330 were previously shown to be substrates for CK1
phosphorylation, a kinase implicated in the regulation of gap junction assembly (Cooper and
Lampe, 2002). Our phosphospecific antibody could react with Cx43 phosphorylated at all 3
sites in heart tissue. We probed heart lysates with 7 other Cx43 phosphospecific antibodies that
we created or are available from commercial sources and that we find to be specific for
phosphorylation at S255, S262, S279/282, S364, S368, or S372. S325/328/330 was the only
one that gave a significant decrease in phosphorylation (results not shown, several showed
little binding to heart lysate under either condition). The antibody to pS368 showed increased
binding upon hypoxia that we have investigated separately (Ek-Vitorin et al., 2006). Similar
to a previous report (Jeyaraman et al., 2003), we found (data not shown) that ischemia led to
increased binding of an anti-Cx43 antibody we prepared to a nonphosphorylated peptide
corresponding to residues 360–382 of Cx43 that appears to be identical to Zymed 13800, an
antibody prepared to a similar epitope that has been reported to bind to “dephosphorylated
Cx43” (Nagy et al., 1997).
We probed the functional consequences of these phosphorylation events by expressing Cx43
with serine to alanine conversions at all 3 sites (i.e., the TM mutant). This mutant, rather than
double or single site mutants, was selected for study for two reasons: 1) all 3 sites appear to be
phosphorylated in heart tissue and 2) the functional consequences of phosphorylation at clusters
of (phospho)serines can be misinterpreted from single site mutants because either the remaining
sites partially or fully compensate for the missing site or the remaining sites are not
phosphorylated because phosphorylation is sequential with the mutated site necessary for
subsequent phosphorylation events (as can be the case for CK1 (Flotow et al., 1990)).
Microinjected Cx43-TM cells and parental (Cx43
−/−
) fibroblasts were occasionally dye
coupled to 1 or 2 cells. In contrast, although electrical coupling between Cx43-TM cells was
Lampe et al. Page 8
J Cell Sci. Author manuscript; available in PMC 2007 February 15.
NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author Manuscript
reduced compared to wild-type in the early hours after plating, significant coupling was
routinely observed at 24–30 hours. This coupling was mediated by channels with conductances
of 70–90pS, a value intermediate to the fully open channel (~110pS) and the previously
reported, phosphorylation-induced 50–60pS substate (Lampe et al., 2000). These results are
consistent with at least two explanations. First, phosphorylation at S325/328/330 is necessary
for Cx43 channels to open to their fully open state. If correct, the intermediate conductance
substate of Cx43 observed in the Cx43-TM cells would have to be poorly permeated by LY to
accommodate the low levels of dye coupling observed for these cells. Interestingly, this
intermediate conductance state is commonly observed in Cx43-WT expressing rat insulinoma
cells that do not express Cx45 (Ek-Vitorin, personal communication), and the junctions formed
by these cells are poorly permeated by dye. Second, since Cx45 is expressed at low levels in
the parental fibroblasts (Martyn et al., 1997), these intermediate channels could reflect the
activity of heteromeric Cx43–Cx45 channels (Martinez et al., 2002). This seems unlikely for
two reasons, first both single channel data sets (Cx43-TM and Cx43-WT) were obtained from
cells whose macroscopic levels of coupling were comparable, and second heteromeric Cx43–
Cx45 channels are not permeated by Lucifer Yellow (Martinez et al., 2002). Regardless of
which of these possibilities is correct, the data indicate that loss of the capacity for
phosphorylation at S325, S328 and/or S330 renders the junctions inefficient at dye transfer
and less conductive.
What mechanisms might control the level of phosphorylation at S328 and/or 330 during
hypoxia? Several groups have shown that phosphatase inhibitors block the change in Cx43
electrophoretic mobility (Duthe et al., 2001; Jeyaraman et al., 2003; Turner et al., 2004). PP1
has been reported to be the main phosphatase that is active against Cx43 in cardiomyocytes
and it appears to be constitutively active (Duthe et al., 2001; Jeyaraman et al., 2003; Turner et
al., 2004). We have shown that CK1 can phosphorylate Cx43 at S325, 328, and/or 330, and
CK1 inhibition leads to loss of phosphorylated Cx43 supporting the role of this kinase in the
phosphorylation of these sites (Cooper and Lampe, 2002). However, a constant kinase/
phosphatase cycling process has been reported to occur during hypoxia with Cx43
dephosphorylation resulting from reduced cellular ATP levels without kinase or phosphatase
activity changes or Cx43 synthesis (Duthe et al., 2001; Turner et al., 2004). Given that kinase
or phosphatase localization could also be critical, the roles that kinase activity, phosphatase
activity and ATP concentration play on the control of the level of phosphorylation at
S325/328/330 during hypoxia remain to be determined.
In summary, data presented here indicate that Cx43 present in cardiac tissue is phosphorylated
at serines 325, 328 and/or 330. During ischemia, the amount of total Cx43 is reduced at the
intercalated disk and Cx43 relocates to the myocytes’ lateral edges. Cx43 at the intercalated
disk retained phosphorylation at S325/328/330 whereas Cx43 present at the lateral edges did
not. We have shown that phosphorylation at these residues was necessary to form the P2
isoform, and these residues are important in gap junction assembly/stability. Thus, the expected
net effect of ischemia would be a measured reduction in longitudinal conduction velocity, due
to decreased localization at intercalated disks (Poelzing and Rosenbaum, 2004), without a
concomitant increase in transverse conduction velocity, despite increased localization at lateral
borders. Whether reduced phosphorylation at S325/328/330 triggers the relocalization of Cx43
to the lateral edges is not known. However, the spatially specific retention of phosphorylation
at these sites at the intercalated disk and loss at the lateral edges would be expected to help
retain the anisotropy of conduction in the heart during ischemia.
Acknowledgements
These studies were supported by Grants from the National Institutes of Health: GM055632 (PDL) and HL058732
(JMB).
Lampe et al. Page 9
J Cell Sci. Author manuscript; available in PMC 2007 February 15.
NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author Manuscript
References
Bao X, Reuss L, Altenberg GA. Regulation of purified and reconstituted connexin 43 hemichannels by
protein kinase C-mediated phosphorylation of Serine 368. J Biol Chem 2004;279:20058–20066.
[PubMed: 14973142]
Beardslee MA, Lerner DL, Tadros PN, Laing JG, Beyer EC, Yamada KA, Kleber AG, Schuessler RB,
Saffitz JE. Dephosphorylation and intracellular redistribution of ventricular connexin43 during
electrical uncoupling induced by ischemia. Circulation Research 2000;87:656–662. [PubMed:
11029400]
Bergoffen J, Scherer SS, Wang S, Oronzi Scott M, Bone LJ, Paul DL, Chen K, Lensch MW, Chance PF,
Fishbeck KH. Connexin mutations in X-linked Charcot-Marie-Tooth disease. Science
1993;262:2039–2042. [PubMed: 8266101]
Burt JM, Spray DC. Volatile anesthetics block intercellular communication between neonatal rat
myocardial cells. Circ Res 1989;65:829–837. [PubMed: 2766493]
Cooper CD, Lampe PD. Casein kinase 1 regulates connexin43 gap junction assembly. J Biol Chem
2002;277:44962–44968. [PubMed: 12270943]
Cottrell GT, Lin R, Warn-Cramer BJ, Lau AF, Burt JM. Mechanism of v-Src- and mitogen-activated
protein kinase-induced reduction of gap junction communication. Am J Physiol Cell Physiol
2003;284:C511–520. [PubMed: 12388103]
Duthe F, Plaisance I, Sarrouilhe D, Herve JC. Endogenous protein phosphatase 1 runs down gap junctional
communication of rat ventricular myocytes. American Journal of Physiology - Cell Physiology
2001;281:C1648–1656. [PubMed: 11600429]
Ek-Vitorin JF, King TJ, Heyman NS, Lampe PD, Burt JM. Selectivity of Connexin 43 Channels Is
Regulated Through Protein Kinase C-Dependent Phosphorylation. Circ Res. 2006
Flotow H, Graves PR, Wang AQ, Fiol CJ, Roeske RW, Roach PJ. Phosphate groups as substrate
determinants for casein kinase I action. J Biol Chem 1990;265:14264–14269. [PubMed: 2117608]
Gong X, Li E, Klier G, Huang Q, Wu Y, Lei H, Kumar NM, Horwitz J, Gilula NB. Disruption of alpha3
connexin gene leads to proteolysis and cataractogenesis in mice. Cell 1997;91:833–843. [PubMed:
9413992]
Jain SK, Schuessler RB, Saffitz JE. Mechanisms of delayed electrical uncoupling induced by ischemic
preconditioning. Circ Res 2003;92:1138–1144. [PubMed: 12730093]
Jeyaraman M, Tanguy S, Fandrich RR, Lukas A, Kardami E. Ischemia-induced dephosphorylation of
cardiomyocyte connexin-43 is reduced by okadaic acid and calyculin A but not fostriecin. Mol Cell
Biochem 2003;242:129–134. [PubMed: 12619875]
Kelsell DP, Dunlop J, Stevens HP, Lench NJ, Laing JN, Parry G, Mueller RF, Leigh IM. Connexin 26
mutations in hereditary non-syndromic sensorineural deafness. Nature 1997;387:80–83. [PubMed:
9139825]
King TJ, Lampe PD. The gap junction protein connexin32 is a mouse lung tumor suppressor. Cancer Res
2004;64:7191–7196. [PubMed: 15492231]
Laird DL, Castillo M, Kasprzak L. Gap junction turnover, intracellular trafficking, and phosphorylation
of connexin43 in Brefeldin A-treated rat mammary tumor cells. J Cell Biol 1995;131:1193–1203.
[PubMed: 8522583]
Int J of Biochem Cell Biol. 36. 2004. The effects of connexin phosphorylation on gap junctional
communication.
Lampe PD, Kurata WE, Warn-Cramer B, Lau AF. Formation of a distinct connexin43 phosphoisoform
in mitotic cells is dependent upon p34
cdc2
kinase. J Cell Sci 1998;111:833–841. [PubMed: 9472011]
Lampe PD, TenBroek EM, Burt JM, Kurata WE, Johnson RG, Lau AF. Phosphorylation of connexin43
on serine368 by protein kinase C regulates gap junctional communication. J Cell Biol
2000;126:1503–1512. [PubMed: 10871288]
Martinez AD, Hayrapetyan V, Moreno AP, Beyer EC. Connexin43 and connexin45 form heteromeric
gap junction channels in which individual components determine permeability and regulation. Circ
Res 2002;90:1100–1107. [PubMed: 12039800]
Lampe et al. Page 10
J Cell Sci. Author manuscript; available in PMC 2007 February 15.
NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author Manuscript
Martyn KD, Kurata WE, Warn-Cramer BJ, Burt JM, TenBroek E, Lau AF. Immortalized connexin43
knockout cell lines display a subset of biological properties associated with the transformed
phenotype. Cell Growth & Differentiation 1997;8:1015–1027. [PubMed: 9300183]
Musil LS, Goodenough DA. Biochemical analysis of connexin43 intracellular transport, phosphorylation
and assembly into gap junctional plaques. J Cell Biol 1991;115:1357–1374. [PubMed: 1659577]
Nagy JI, Li WEI, Roy C, Doble BW, Gilchrist JS, Kardami E, Hertzberg EL. Selective monoclonal
antibody recognition and cellular localization of an unphosporylated form of connexin43. Exper Cell
Res 1997;236:127–136. [PubMed: 9344592]
Paznekas WA, Boyadjiev SA, Shapiro RE, Daniels O, Wollnik B, Keegan CE, Innis JW, Dinulos MB,
Christian C, Hannibal MC, et al. Connexin 43 (GJA1) mutations cause the pleiotropic phenotype of
oculodentodigital dysplasia. Am J Hum Genet 2003;72:408–418. [PubMed: 12457340]
Poelzing S, Rosenbaum DS. Nature, significance, and mechanisms of electrical heterogeneities in
ventricle. Anat Rec A Discov Mol Cell Evol Biol 2004;280:1010–1017. [PubMed: 15368342]
Saez JC, Berthoud VM, Branes MC, Martinez AD, Beyer EC. Plasma membrane channels formed by
connexins: their regulation and functions. Physiol Rev 2003;83:1359–1400. [PubMed: 14506308]
Schulz R, Gres P, Skyschally A, Duschin A, Belosjorow S, Konietzka I, Heusch G. Ischemic
preconditioning preserves connexin 43 phosphorylation during sustained ischemia in pig hearts in
vivo. Faseb J 2003;17:1355–1357. [PubMed: 12759340]
Severs NJ, Dupont E, Coppen SR, Halliday D, Inett E, Baylis D, Rothery S. Remodelling of gap junctions
and connexin expression in heart disease. Biochim Biophys Acta 2004;1662:138–148. [PubMed:
15033584]
Sohl G, Willecke K. Gap junctions and the connexin protein family. Cardiovasc Res 2004;62:228–232.
[PubMed: 15094343]
Solan JL, Fry MD, TenBroek EM, Lampe PD. Connexin43 phosphorylation at S368 is acute during S
and G2/M and in response to protein kinase C activation. J Cell Sci 2003;116:2203–2211. [PubMed:
12697837]
TenBroek EM, Lampe PD, Solan JL, Reynhout JK, Johnson RG. Ser364 of connexin43 and the
upregulation of gap junction assembly by cAMP. J Cell Biol 2001;155:1307–1318. [PubMed:
11756479]
Turner MS, Haywood GA, Andreka P, You L, Martin PE, Evans WH, Webster KA, Bishopric NH.
Reversible connexin 43 dephosphorylation during hypoxia and reoxygenation is linked to cellular
ATP levels. Circ Res 2004;95:726–733. [PubMed: 15358666]
White T, Paul D. Genetic diseases and gene knockouts reveal diverse connexin functions. Ann Rev
Physiology 1999;61:283–310.
Lampe et al. Page 11
J Cell Sci. Author manuscript; available in PMC 2007 February 15.
NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author Manuscript
Fig. 1.
Increased electrophoretic mobility of Cx43 in the ischemic heart involves dephosphorylation
at serines 325/328/330. Western blot of total protein isolated from 3 control (Lanes 1,3,5,
respectively) or 3 globally ischemic (30’ at 37°C; Lanes 2,4,6, respectively) hearts
simultaneously probed for total Cx43 (Cx43NT-1) and Cx43 phosphorylated at S325/328/330
followed by GAPDH, and vinculin antibodies. Note the increased mobility of Cx43 and
decreased pS325/328/330 content in the ischemic samples. Molecular weight standards are 28,
49, 62 and 98 kDa.
Lampe et al. Page 12
J Cell Sci. Author manuscript; available in PMC 2007 February 15.
NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author Manuscript
Fig. 2.
Differential phosphorylation of Cx43 localized at intercalated disks vs. lateral edges of
myocytes in the normal and ischemic heart. The localization of total Cx43 (Sigma C6219) and
pS325/328/330-Cx43 in normal and ischemic heart is shown. Note the increase in Cx43 at the
lateral edges of the myocytes of the ischemic heart. Lateral edge Cx43 was not detectably
phosphorylated at S325/328/330. In contrast, Cx43 at the intercalated disks of ischemic heart
retained phosphorylation at S325/328/330 (bar=20μm).
Lampe et al. Page 13
J Cell Sci. Author manuscript; available in PMC 2007 February 15.
NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author Manuscript
Fig. 3.
The pS325/328/330 antibody is specific for the P2 isoform of Cx43 and pS328/330. (A) Equal
amounts of protein lysate from fibroblasts expressing wild-type Cx43 (WT) and Cx43-TM
(TM, clone A) were either treated with alkaline phosphatase (+AP) or untreated and probed in
a Western blot with Cx43NT1 (Total Cx43-left panel) and the pS325/328/330 (right panel)
antibodies on the same blot. Molecular weight standards are marked on the left side and the
P2 isoform is marked with an *. (B) The amount of antibody bound to the different peptides
representing singly, doubly and triply phosphorylated Cx43 linked to an ELISA well (open
bars) is shown along with the ability of these peptides to compete with Cx43 present in heart
lysates in a Western immunoblot format (filled bars). Error bars represent standard deviation
from the mean.
Lampe et al. Page 14
J Cell Sci. Author manuscript; available in PMC 2007 February 15.
NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author Manuscript
Fig. 4.
Cx43-TM expressing cells do not phosphorylate Cx43 to the P2 isoform and contain less Triton
insoluble (TI) Cx43. Whole cell (WC) and Triton X-100 insoluble (TI) cellular lysates from
cells expressing wild-type Cx43 (WT) or three Cx43-TM clones (TM-A, TM-B, and TM-C)
were western immunoblotted and probed for total Cx43 with the Cx43NT1 antibody. A darker
exposure of the TM-C clone is shown. The migration position of a 50 and 36 kDa marker are
indicated with an * and the vinculin loading controls are indicated with a V. The Triton
insoluble vinculin could be associated with the cytoskeleton or adhesive junctions.
Lampe et al. Page 15
J Cell Sci. Author manuscript; available in PMC 2007 February 15.
NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author Manuscript
Fig. 5.
Comparison of Cx43 localization in parental knockout (KO) cells, wild-type Cx43 (WT) and
Cx43-TM (TM-A) expressing cells. Apparent gap junctions are marked by arrowheads in the
WT panel. Bar=10μm.
Lampe et al. Page 16
J Cell Sci. Author manuscript; available in PMC 2007 February 15.
NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author Manuscript
Fig. 6.
Cells expressing Cx43 S325/328/330A are inefficient at dye transfer. Wild-type (WT), parental
(KO) and Cx43-TM (clones TM-A, TM-B and TM-C) expressing cells were microinjected
with Lucifer yellow dye. After 3 minutes of transfer, digital images were taken and the number
of recipient cells was determined (n=the number of injected cells). Mean and standard error
are shown. The extent of dye transfer was significantly different (p<0.02) for WT compared
to all of the other cell types as indicated by **.
Lampe et al. Page 17
J Cell Sci. Author manuscript; available in PMC 2007 February 15.
NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author Manuscript
Fig. 7.
Histogram of channel events observed in Cx43-TM (gray) and Cx43-WT expressing cells.
Panel A shows event frequency as a percent of total events in each bin; panel B shows the
difference induced by mutation of the 325, 328, and 330 sites. Note the reduced incidence of
events corresponding to fully open channels (106 ± 1 pS) and increased incidence of 75 ± 2
pS. The 54 ± 1 pS population did not differ between wild-type and mutant data sets. Difference
data were fit using Origin, Chi
2
and R
2
for fit are 5.677 and 0.94, respectively (−WT: 1851
events, n=13; TM: 418 events, n=7).
Lampe et al. Page 18
J Cell Sci. Author manuscript; available in PMC 2007 February 15.
NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author Manuscript
