Although recent studies have revealed that heart cells are generated in adult mammals,
the frequency and source of new heart cells is unclear. Some studies suggest a high
rate of stem cell activity with differentiation of progenitors to cardiomyocytes
1
. Other studies suggest that new cardiomyocytes are born at a very low rate
2,3,4
, and that they may be derived from division of pre-existing cardiomyocytes. Thus,
the origin of cardiomyocytes in adult mammals remains unknown. Here we combined two
different pulse-chase approaches -- genetic fate-mapping with stable isotope labeling
and Multi-isotope Imaging Mass Spectrometry (MIMS). We show that genesis of cardiomyocytes
occurs at a low rate by division of pre-existing cardiomyocytes during normal aging,
a process that increases by four-fold adjacent to areas of myocardial injury. Cell
cycle activity during normal aging and after injury led to polyploidy and multinucleation,
but also to new diploid, mononucleated cardiomyocytes. These data reveal pre-existing
cardiomyocytes as the dominant source of cardiomyocyte replacement in normal mammalian
myocardial homeostasis as well as after myocardial injury.
Despite intensive research, fundamental aspects of the mammalian heart’s capacity
for self-renewal are actively debated
5,6
. Estimates of cardiomyocyte turnover range from less than 1% per year
2,3,4
to more than 40% per year
7
. Turnover has been reported to either decrease
3
or increase with age
7
, while the source of new cardiomyocytes has been attributed to both division of existing
myocytes
8
and to progenitors residing within the heart
9
or in exogenous niches such as bone marrow
10
. Controversy persists regarding the plasticity of the adult heart in part due to
methodological challenges associated with studying slowly replenished tissues. Toxicity
attributed to radiolabeled thymidine
11
and halogenated nucleotide analogues
12
limits the duration of labeling and may produce direct biological effects. Tissue
autofluorescence can reduce the sensitivity and specificity of immunofluorescent methods
of detecting cell cycle activity
5,13
, including as cell cycle markers or incorporation of halogenated nucleotide analogues.
The challenge of measuring cardiomyocyte turnover is further compounded by the faster
rate of turnover of cardiac stromal cells relative to cardiomyocytes
14
.
We used Multi-isotope Imaging Mass Spectrometry (MIMS) to study cardiomyocyte turnover
and to determine whether new cardiomyocytes are derived from preexisting myocytes
or from a progenitor pool (Fig 1a). MIMS uses ion microscopy and mass spectrometry
to generate high resolution quantitative mass images and localize stable isotope reporters
in domains smaller than one micron cubed
15,16,17
. MIMS generates 14N quantitative mass images by measuring the atomic composition
of the sample surface with a lateral resolution of under 50nm and a depth resolution
of a few atomic layers. Cardiomyocyte cell borders and intracellular organelles were
easily resolved (Fig 1b). Regions of interest can be analyzed at higher resolution,
demonstrating cardiomyocyte-specific subcellular ultrastructure, including sarcomeres
(Fig 1c, Supplemental Fig 1a). In all subsequent analyses, cardiomyocyte nuclei were
identified by their location within sarcomere-containing cells, distinguishing them
from adjacent stromal cells.
An immense advantage of MIMS is the detection of nonradioactive stable isotope tracers.
As an integral part of animate and inanimate matter, they do not alter biochemical
reactions and are not harmful to the organism
18
. MIMS localizes stable isotope tracers by simultaneously quantifying multiple masses
from each analyzed domain; this enables the generation of a quantitative ratio image
of two stable isotopes of the same element
15
. The incorporation of a tracer tagged with the rare stable isotope of nitrogen (15N)
is detectable with high precision by an increase in 15N:14N above the natural ratio
(0.37%). Nuclear incorporation of 15N-thymidine is evident in cells having divided
during a one-week labeling period, as observed in the small intestinal epithelium,
which turns over completely in 3–5 days
16
(Fig 1d); in contrast, 15N-thymidine labeled cells are rarely observed in the heart
(Fig 1e) after 1 week of labeling. In subsequent studies, small intestine was used
as a positive control to confirm label delivery.
To evaluate for an age-related change in cell cycle activity, we administered 15N-thymidine
for 8 weeks to three age groups of C57BL6 mice starting at day-4 (neonate), at 10-weeks
(young adult) and at 22-months (old adult) (Supplemental Fig 2). We then performed
MIMS analysis (Fig 2a, b, Supplemental Fig 3). In the neonatal group, 56% (±3% s.e.m.,
n=3 mice) of cardiomyocytes demonstrated 15N nuclear labeling, consistent with the
well-accepted occurrence of cardiomyocyte DNA synthesis during post-natal development
19
. We observed a marked decline in the frequency of 15N-labeled cardiomyocyte nuclei
(15N+CM) in the young adult (neonate= 1.00%15N+CM/day ±0.05 s.e.m. vs young adult=0.015%
15N+CM/day ±0.003 s.e.m., n=3 mice/group, p<0.001) (Fig 2a, c; Supplemental Fig 3).
We found a further reduction in cardiomyocyte DNA synthesis in old mice (young adult=0.015%15N+CM/day
±0.003 s.e.m. vs. old adult=0.007 %15N+CM/day ±0.002 s.e.m., n=3/group, p<0.05) (Fig
2c). The observed pattern of nuclear 15N-labeling in cardiomyocytes is consistent
with the known chromatin distribution pattern in cardiomyocytes
20
(Supplemental Fig 1b) and was measured at levels that could not be explained by DNA
repair (Supplemental Fig 4). Extrapolating DNA synthesis measured in cardiomyocytes
over 8 weeks yields a yearly rate of 5.5% in the young adult and 2.6% in the old mice.
Given that cardiomyocytes are known to undergo DNA replication without completing
the cell cycle
19,21,22
, these calculations represent the upper limit of cardiomyocyte generation under normal
homeostatic conditions, indicating a low rate of cardiogenesis.
To test whether cell cycle activity occurred in preexisting cardiomyocytes or was
dependent on a progenitor pool, we performed 15N-thymidine labeling of double-transgenic
MerCreMer/ZEG mice, previously developed for genetic lineage mapping (Fig 3a)
23,24
. MerCreMer/ZEG cardiomyocytes irreversibly express green fluorescent protein (GFP)
after treatment with 4OH-tamoxifen, allowing pulse labeling of existing cardiomyocytes
with a reproducible efficiency of approximately 80%. Although some have reported rare
GFP expression by non-cardiomyocytes with this approach
25
, we did not detect GFP expression in interstitial cells isolated from MerCreMer/ZEG
hearts nor did we detect GFP expression by Sca1 or ckit-expressing progenitors in
histological sections (Supplemental Fig 5). Thus, during a chase period, cardiomyocytes
generated from progenitors should be GFP−, whereas cardiomyocytes arising from preexisting
cardiomyocytes should express GFP at a frequency similar to the background rate induced
by 4OH-tamoxifen. We administered 4OH-tamoxifen for two weeks to 8 wk-old mice (n=4);
during a subsequent 10-week chase, mice received 15N-thymidine via osmotic minipump
(20μg/h). With MIMS analysis, we identified 35 15N-labeled cardiomyocyte nuclei (of
4190 analyzed) over 10 weeks, yielding a projected yearly rate of cardiomyocyte DNA
replication of 4.4%. Extrapolating from prior reports of high stem cell-dependent
cardiomyocyte turnover
7
, we had anticipated detecting >320 cardiomyocytes entering the cell cycle; these
results exclude such a high rate of turnover (expected 15N+ cardiomyocytes=321; observed=35;
Fisher’s exact<0.0001). Immunofluorescent staining for GFP was performed on adjacent
sections and an observer unaware of MIMS analysis results assessed GFP status. Of
15N+ cardiomyocytes, 77% expressed GFP, a frequency essentially identical to that
of surrounding 15N− cardiomyocytes (15N+ CM, 77% vs 15N− cardiomyocytes, 84%; Fisher’s
exact=n.s.) (Table 1). If new cardiomyocytes were derived from progenitors, 15N+ cardiomyocytes
would have been GFP− (expected=0/35; observed=27/35, Fisher’s exact<0.0001). These
data reveal that 15N-labeled cardiomyocytes resulted from DNA synthesis by preexisting
cardiomyocytes and exclude a substantial contribution from stem cells to cardiomyocyte
replacement in the uninjured heart.
Cardiomyocytes can undergo DNA replication without completing the cell cycle. Although
multinucleation and polyploidization occur during early post-natal development and
in response to myocardial stress
2,19
, we considered the possibility these processes could account for 15N+ cardiomyocytes
in the uninjured adult mouse. We performed fluorescent in situ hybridization in adjacent
sections to assess the ploidy state of each 15N+ cardiomyocyte and surrounding 15N−
cardiomyocytes, and an observer unaware of the results of MIMS analysis identified
fluorescently-labeled chromosomes (Supplemental Fig 6). Although we found 15N+ cardiomyocytes
that were polyploid (4n or greater), we observed a higher frequency of diploid nuclei
in the 15N+ pool compared to surrounding 15N− cardiomyocytes (15N+ diploid:polyploid
= 22:12 vs 15N− diploid:polyploid = 9:56, Fisher’s exact p<0.0001), consistent with
ongoing cell division. We then assessed each cell for multinucleation using serial
0.5μm sections adjacent to the section used for MIMS analysis (Supplemental Fig 6).
We observed that 49% of 15N+ cardiomyocytes were mononucleated, in contrast to a frequency
of 24% for surrounding 15N− cardiomyocytes (Fisher’s exact p<0.01), also consistent
with cell division. The majority of cardiomyocyte DNA synthesis occurred in polyploid
and/or multinucleated cardiomyocytes as might be expected with a physiologic hypertrophic
response and thus unlikely to indicate cardiomyocyte division; however, 17% (6 of
35 15N+ CM) were diploid and mononucleated, consistent with newly generated cardiomyocytes
(Supplemental Fig 7). The mononucleated, diploid, 15N+ cardiomyocytes were also predominantly
GFP+ (5 of 6 = 83% vs. 82% background frequency, p=n.s.), suggesting that they arose
from preexisting cardiomyocytes at a slow annual rate of 0.76% (n=6 of 4190 over 10
weeks).
We next used MIMS and genetic fate mapping to study myocardial injury. Cardiomyocyte
GFP labeling was induced in MerCreMer/ZEG mice with 4OH-tamoxifen. Mice then underwent
experimental myocardial infarction or sham surgery followed by continuous labeling
with 15N-thymidine for 8wks. The frequency of 15N-labeled cardiomyocytes in sham-operated
mice was similar to prior experiments in unoperated mice (yearly projected rates:
sham=6.8%; unoperated=4.4%), but increased significantly adjacent to infarcted myocardium
(total 15N+ cardiomyocyte nuclei: MI=23.0% vs sham=1.1%, Fig 4a–b, Supplemental Fig
8). We examined GFP expression, nucleation and ploidy status of 15N-labeled cardiomyocytes
and surrounding unlabeled cardiomyocytes. We found a significant dilution of the GFP+
cardiomyocyte pool at the border region as previously shown
23,24
(67% vs. 79%, p<0.05, Table 2, Supplemental Fig 9); however, 15N+ myocytes demonstrated
a similar frequency of GFP expression compared to unlabeled myocytes (71% vs. 67%,
Fisher’s exact=n.s.), suggesting that DNA synthesis was primarily occurring in pre-existing
cardiomyocytes. Of 15N-labeled cardiomyocytes, approximately 14% were mononucleated
and diploid consistent with division of pre-existing cardiomyocytes (Supplemental
Fig 6, 7). We observed higher DNA content (>2N) in the remaining cardiomyocytes as
expected with compensatory hypertrophy after injury. Thus, in the 8wks after myocardial
infarction, approximately 3.2% of the cardiomyocytes adjacent to the infarct had unambiguously
undergone division (total 15N+ × mononucleated diploid fraction = 23% × 0.14 = 3.2%).
The low rate of cardiomyocyte cell cycle completion is further supported by the absence
of detectable Aurora B Kinase, a transiently expressed cytokinesis marker, which was
detected in rapidly proliferating small intestinal cells but not in cardiomyocytes
(Supplemental Fig 10). We also considered the possibility that a subset of 15N+ myocytes
that were multinucleated and/or polyploid resulted from division followed by additional
rounds of DNA synthesis without division. However, quantitative analysis of the 15N+
population did not identify a subpopulation that had accumulated additional 15N-label
as would be expected in such a scenario (Supplemental Fig 11). Together, these data
suggest that adult cardiomyocytes retain some capacity to reenter the cell cycle,
but that the majority of DNA synthesis after injury occurs in preexisting cardiomyocytes
without completion of cell division.
If dilution of the GFP+ cardiomyocyte pool cannot be attributed to division and differentiation
of endogenous progenitors, do these data exclude a role for progenitors in the adult
mammalian heart? These data could be explained by preferential loss of GFP+ cardiomyocytes
after injury, a process that we have previously considered but for which we have not
found supporting evidence
23
. Such an explanation excludes a role for endogenous progenitors in cardiac repair
and would be consistent with data emerging from lower vertebrates
8,26
and the neonatal mouse
27
in which preexisting cardiomyocytes are the cellular source for cardiomyocyte repletion.
A second possibility to explain the dilution of the GFP+ cardiomyocyte pool is that
injury stimulates progenitor differentiation without division; inevitably, this would
lead to exhaustion of the progenitor pool, which if true could explain the limited
regenerative potential of the adult mammalian heart.
In summary, this study demonstrates birth of cardiomyocytes from preexisting cardiomyocytes
at a projected rate of approximately 0.76%/year (15N+ annual rate × mononucleated
diploid fraction = 4.4% × 0.17) in the young adult mouse under normal homeostatic
conditions, a rate that declines with age but increases by approximately four-fold
after myocardial injury in the border region. This study shows that cardiac progenitors
do not play a significant role in myocardial homeostasis in mammals and suggests that
their role after injury is also limited.
Online Methods
Mice
All experiments were conducted in accordance with the Guide for the Use and Care of
Laboratory Animals and approved by the Harvard Medical School Standing Committee on
Animals. C57Bl/6 male mice were obtained from Charles River.
We generated double transgenic Mer-CreMer-ZEG male mice by crossbreeding cardiomyocyte-specific
MerCreMer mice and ZEG mice (Jackson Laboratory). β-galactosidase-GFP is under the
control of a cytomegalovirus (CMV) enhancer/chicken actin promoter (Actb); the background
strain was C57BL/6J (N7). The background strain of the MerCreMer mice was C57Bl/6SV129.
Genotyping was performed by PCR on tail DNA using the following primers: MerCreMer
forward: 5′-GTCTGAC TAGGTGTCCTTCT-3′; MerCreMer backward: 5′-CGTCCTCCTGCTGGTA TAG-3′;
ZEG forward: 5′-AAGTTCATCTGCACCACCG-3′; ZEG backward: 5′-TCCTTGAAGAAGATGGTGCG-3′;
ZEG control forward: 5′-CTAGGCCA CAGAATTGAAAGATCT-3′; and ZEG control backward: 5′-GTAGGTG
GAAATTCTAGCATCATCC-3′. To induce Cre recombination and GFP labeling in cardiomyocytes,
we injected 4-OH-tamoxifen (provided by a generous gift from Laboratoires Besins),
dissolved in peanut oil (Sigma) at a concentration of 5 mg/ml, intraperitoneally into
8-week-old MerCreMer-ZEG mice daily at a dosage of 0.5 mg/day for 14 days.
Experimental myocardial infarction
Mice were subjected to experimental myocardial infarction as described. Surgeries
were performed by a single operator with more than 20 years of experience in the performance
of coronary ligation in rodents. In brief, the left coronary artery was permanently
ligated approximately 2 mm below the left atrial appendage. For sham operations, the
thoracic cavity was opened and the heart exposed, but no intramyocardial sutures were
placed. 15N-thymidine (Cambridge Isotopes) was administered at a rate of 20μg/hr via
osmotic minipump (Alzet), implanted subcutaneously at the time of experimental myocardial
infarction after a single intraperitoneal bolus dose of 500μg.
MIMS data acquisition
The factory prototype of the NanoSIMS50 as well as a standard NanoSIMS 50 and a large
radius NanoSIMS 50L (Cameca, Gennevilliers, France) was used for MIMS analysis as
previously described
15
. A focused beam of Cs+ ions was used to sputter a few atomic layers and generate
secondary ions from the left ventricular free wall of heart section samples. The Cs+
primary ions were scanned over a raster pattern of either 256 × 256 pixels or 512
× 512 pixels. At each pixel location, the secondary ion intensities for 12C−, 13C−,
12C14N− and 12C15N−were recorded in parallel from the same sputtered volume. The detection
of nitrogen requires the use of cluster ions 12C14N− and 12C15N− for 14N and 15N,
respectively, due to the low ionization efficiency of nitrogen as N−.
MIMS data analysis
From a single field image acquisition, we first extracted four image files: the four
original quantitative mass images (QMIs; 12C, 13C, 12C14N, 12C15N) and the two ratio
images (13C/12C and 12C15N/12C14N), derived from the pixel-by-pixel division of the
13C QMI by the 12C QMI and of the 12C15N QMI by the 12C14N QMI, respectively. We then
used a hue saturation intensity transformation (HSI) of the ratio image to map 15N-labeled
regions. The hue corresponds to the ratio value, and the intensity at a given hue
is an index of statistical reliability.
15N-thymidine labeling
For the neonatal cohort in the aging experiment, labeling was starting at post-natal
day 4 with subcutaneous injections of 50μg/g 15N-thymidine (Cambridge Isotopes) every
12h and continued through post-natal week 4. Starting at age 4 wks – and in all other
experiments using adult mice – osmotic minipumps (Alzet) were implanted subcutaneously,
delivering 15N-thymidine (Cambridge Isotopes) at a rate of 20 μg/hour.
Multinucleation Analysis
Serial adjacent sections (0.5μm) were stained to identify cardiomyocyte borders. A
given cardiomyocyte was tracked in the vertical axis by locating it in serial sections.
Uninjured hearts were stained using a modified PAS protocol with standard solutions
(Electron Microscopy Services), but with longer incubation times optimized for LR
white embedding. Slides were incubated in xylene at 37C, 1 hour, rehydrated through
graded alcohols, incubated in periodic acid for two hours, and Schiff’s reagent for
two nights. Sections were counterstained in hematoxylin and Scott’s Bluing for 1 hour
each. Injured hearts were stained using a modified Trichrome staining protocol with
standard solutions (Fisher Scientific), but with longer incubation times. Slides were
incubated in xylene at 37C, 1 hour, rehydrated through graded alcohols, incubated
in bouin’s fluid at 56C, 1 hour, rinsed in tap water, incubated in weigert’s iron
hematoxylin stain 1 hour, rinsed in tap water, incubated in scarlet-acid fuchsin solution,
1 hour, rinsed in DI water, incubated in phophotungistic-phophomolybdic acid solution,
30 minutes, incubated in aniline blue stain solution, 30 minutes, and incubated in
1% acetic acid, 20 minutes.
Fluorescent in situ hybridization
Sections were incubated in proteinase K (50ug/ml) at 60C for 15min. After a PBS/MgCl2
(45mM) wash, slides were post-fixed in 4% paraformaldehyde (PBS/MgCl2), dehydrated
through graded alcohols. Biotinylated-labeled chromosome Y paint (Star-FISH, Cambio)
in hybridization mix was applied to sections, and sealed under glass with rubber cement
(note some samples were analyzed with chromosome-18 paint due to product discontinuation
of Y-paint). Samples were heated to 90C for 15 min. After an overnight incubation
at 37C, slides were washed three times with 50% formamide/2x standard saline citrate
at 45C, three washes with 2x standard saline citrate at room temperature, and two
washes with 4x standard saline citrate/0.1% Tween at room temperature. Samples were
blocked 10 minutes with 4x standard saline citrate/0.1% Tween/ 0.05% milk and incubated
for 2 hr in streptavidin-conjugated Alexa Fluor 488 (Invitrogen) prior to washing
and mounting. An observer unaware of the MIMS images or 15N-thymidine labeling status
of the nuclei assigned ploidy status.
Immunofluorescent staining
Sections were incubated in glycine/tris (50mM glycine/0.05M Tris) at room temperature
for 5min. After a brief wash with Tris, sections were incubated with chicken anti-GFP
(Abcam), rabbit anti-ckit (Abcam), rat anti-sca1 (Abcam) with fresh 0.1%BSA in TBST
(TBS/0.1%Tween) overnight at 4C. After a brief wash with TBS, sections were incubated
with anti-chicken Alexa Fluor 488 (Invitrogen) prior to TBS wash and mounting. An
observer unaware of the MIMS images or 15N-thymidine labeling status of the nuclei
assigned GFP status.
Fluorescence Image analysis
We used a custom-written script in IP Lab version 4.0 (Scanalytics) imaging software
for serial image acquisition. Tissue sections were auto-imaged using an Olympus IX-70
microscope with a digital charge-coupled device camera (CoolSNAP EZ, Roper Scientific),
an automated stage with a piezoelectric objective positioner (Polytec PI, Auburn MA)
to compensate for deviations in the z-axis. Images were compressed and stitched into
a mosaic using stitching software (Canon Photostitch). Multichannel images were merged
in ImageJ prior to stitching.
Statistical analysis
Statistical testing was performed using Prism 3.0 (Graphpad). Results are presented
as mean ± s.e.m. and were compared using T-tests (significance was assigned for p<0.05).
Data comparing event rates were tested with a Fisher-Exact test.
Supplementary Material
1